[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13772934#comment-13772934 ] Hudson commented on YARN-353: - SUCCESS: Integrated in Hadoop-Yarn-trunk #338 (See [https://builds.apache.org/job/Hadoop-Yarn-trunk/338/]) YARN-353. Add Zookeeper-based store implementation for RMStateStore. Contributed by Bikas Saha, Jian He and Karthik Kambatla. (hitesh: http://svn.apache.org/viewcvs.cgi/?root=Apache-SVNview=revrev=1524829) * /hadoop/common/trunk/hadoop-common-project/hadoop-common/src/test/java/org/apache/hadoop/ha/ClientBaseWithFixes.java * /hadoop/common/trunk/hadoop-yarn-project/CHANGES.txt * /hadoop/common/trunk/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-api/src/main/java/org/apache/hadoop/yarn/conf/YarnConfiguration.java * /hadoop/common/trunk/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-common/src/main/resources/yarn-default.xml * /hadoop/common/trunk/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/pom.xml * /hadoop/common/trunk/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/FileSystemRMStateStore.java * /hadoop/common/trunk/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/RMStateStore.java * /hadoop/common/trunk/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/ZKRMStateStore.java * /hadoop/common/trunk/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/TestRMStateStore.java * /hadoop/common/trunk/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/TestZKRMStateStoreZKClientConnections.java Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Karthik Kambatla Fix For: 2.3.0 Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.12.patch, yarn-353-12-wip.patch, YARN-353.13.patch, YARN-353.14.patch, YARN-353.15.patch, YARN-353.16.1.patch, YARN-353.16.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13771670#comment-13771670 ] Hadoop QA commented on YARN-353: {color:green}+1 overall{color}. Here are the results of testing the latest attachment http://issues.apache.org/jira/secure/attachment/12603992/YARN-353.16.1.patch against trunk revision . {color:green}+1 @author{color}. The patch does not contain any @author tags. {color:green}+1 tests included{color}. The patch appears to include 3 new or modified test files. {color:green}+1 javac{color}. The applied patch does not increase the total number of javac compiler warnings. {color:green}+1 javadoc{color}. The javadoc tool did not generate any warning messages. {color:green}+1 eclipse:eclipse{color}. The patch built with eclipse:eclipse. {color:green}+1 findbugs{color}. The patch does not introduce any new Findbugs (version 1.3.9) warnings. {color:green}+1 release audit{color}. The applied patch does not increase the total number of release audit warnings. {color:green}+1 core tests{color}. The patch passed unit tests in hadoop-common-project/hadoop-common hadoop-yarn-project/hadoop-yarn/hadoop-yarn-api hadoop-yarn-project/hadoop-yarn/hadoop-yarn-common hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager. {color:green}+1 contrib tests{color}. The patch passed contrib unit tests. Test results: https://builds.apache.org/job/PreCommit-YARN-Build/1969//testReport/ Console output: https://builds.apache.org/job/PreCommit-YARN-Build/1969//console This message is automatically generated. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Karthik Kambatla Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.12.patch, yarn-353-12-wip.patch, YARN-353.13.patch, YARN-353.14.patch, YARN-353.15.patch, YARN-353.16.1.patch, YARN-353.16.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13771956#comment-13771956 ] Karthik Kambatla commented on YARN-353: --- Thanks Hitesh. Dropping the try-catch blocks seems reasonable to me. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Karthik Kambatla Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.12.patch, yarn-353-12-wip.patch, YARN-353.13.patch, YARN-353.14.patch, YARN-353.15.patch, YARN-353.16.1.patch, YARN-353.16.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13772275#comment-13772275 ] Bikas Saha commented on YARN-353: - We should be storing the information inside the data of the znode instead of encoding it in the name of the znode. The rename pattern was followed as an optimization in FSStore but is not necessary in ZKStore give set-with-overwrite operations. Its also better to avoid encoding the info in the name since doing ls in ZK is more relaxed wrt permissions than it is in FileSystem. {code} + if (childNodeName.startsWith(DELEGATION_TOKEN_SEQUENCE_NUMBER_PREFIX)) { +rmState.rmSecretManagerState.dtSequenceNumber = +Integer.parseInt(childNodeName.split(_)[1]) {code} We should move to creating a node hierarchy for apps such that all znodes for an app are stored under an app znode instead of the app root znode. This will help in removeApplication and also in scaling better on ZK. The earlier code was written this way to ensure create/delete happens under a root znode for fencing. But given that we have moved to multi-operations globally, this isnt required anymore. storeRMDTMasterKeyState() is not using multi-operation. We should be issuing every action on ZK using a multi-operation for fencing to work. I am fine with the patch as is since it puts the framework in place. The above comments may be addressed in YARN-1222 thats opened under YARN-149 for HA. YARN-1222 blocks YARN-1026. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Karthik Kambatla Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.12.patch, yarn-353-12-wip.patch, YARN-353.13.patch, YARN-353.14.patch, YARN-353.15.patch, YARN-353.16.1.patch, YARN-353.16.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13772306#comment-13772306 ] Arun C Murthy commented on YARN-353: No, I think we are good. branch-2.1 is too close for major additions. Thanks. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Karthik Kambatla Fix For: 2.3.0 Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.12.patch, yarn-353-12-wip.patch, YARN-353.13.patch, YARN-353.14.patch, YARN-353.15.patch, YARN-353.16.1.patch, YARN-353.16.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13772296#comment-13772296 ] Hudson commented on YARN-353: - SUCCESS: Integrated in Hadoop-trunk-Commit #4443 (See [https://builds.apache.org/job/Hadoop-trunk-Commit/4443/]) YARN-353. Add Zookeeper-based store implementation for RMStateStore. Contributed by Bikas Saha, Jian He and Karthik Kambatla. (hitesh: http://svn.apache.org/viewcvs.cgi/?root=Apache-SVNview=revrev=1524829) * /hadoop/common/trunk/hadoop-common-project/hadoop-common/src/test/java/org/apache/hadoop/ha/ClientBaseWithFixes.java * /hadoop/common/trunk/hadoop-yarn-project/CHANGES.txt * /hadoop/common/trunk/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-api/src/main/java/org/apache/hadoop/yarn/conf/YarnConfiguration.java * /hadoop/common/trunk/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-common/src/main/resources/yarn-default.xml * /hadoop/common/trunk/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/pom.xml * /hadoop/common/trunk/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/FileSystemRMStateStore.java * /hadoop/common/trunk/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/RMStateStore.java * /hadoop/common/trunk/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/ZKRMStateStore.java * /hadoop/common/trunk/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/TestRMStateStore.java * /hadoop/common/trunk/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/TestZKRMStateStoreZKClientConnections.java Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Karthik Kambatla Fix For: 2.3.0 Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.12.patch, yarn-353-12-wip.patch, YARN-353.13.patch, YARN-353.14.patch, YARN-353.15.patch, YARN-353.16.1.patch, YARN-353.16.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13771393#comment-13771393 ] Karthik Kambatla commented on YARN-353: --- [~hitesh], can you please take a look when you get a chance? I believe the latest patch addresses all the comments. Thanks. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Karthik Kambatla Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.12.patch, yarn-353-12-wip.patch, YARN-353.13.patch, YARN-353.14.patch, YARN-353.15.patch, YARN-353.16.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13768725#comment-13768725 ] Karthik Kambatla commented on YARN-353: --- The latest patch addresses Hitesh's latest set of review comments. Some of the spurious looking changes in TestRMRestart or to fix lines longer than 80 chars. I tested this on a pseudo-dist cluster: # Started the RM using ZKRMStateStore # Ran jobs - all ZKnodes are created as expected # Restarted the RM several times while a job is running - the job succeeds. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Karthik Kambatla Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.12.patch, yarn-353-12-wip.patch, YARN-353.13.patch, YARN-353.14.patch, YARN-353.15.patch, YARN-353.16.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13768773#comment-13768773 ] Hadoop QA commented on YARN-353: {color:green}+1 overall{color}. Here are the results of testing the latest attachment http://issues.apache.org/jira/secure/attachment/12603419/YARN-353.16.patch against trunk revision . {color:green}+1 @author{color}. The patch does not contain any @author tags. {color:green}+1 tests included{color}. The patch appears to include 3 new or modified test files. {color:green}+1 javac{color}. The applied patch does not increase the total number of javac compiler warnings. {color:green}+1 javadoc{color}. The javadoc tool did not generate any warning messages. {color:green}+1 eclipse:eclipse{color}. The patch built with eclipse:eclipse. {color:green}+1 findbugs{color}. The patch does not introduce any new Findbugs (version 1.3.9) warnings. {color:green}+1 release audit{color}. The applied patch does not increase the total number of release audit warnings. {color:green}+1 core tests{color}. The patch passed unit tests in hadoop-common-project/hadoop-common hadoop-yarn-project/hadoop-yarn/hadoop-yarn-api hadoop-yarn-project/hadoop-yarn/hadoop-yarn-common hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager. {color:green}+1 contrib tests{color}. The patch passed contrib unit tests. Test results: https://builds.apache.org/job/PreCommit-YARN-Build/1939//testReport/ Console output: https://builds.apache.org/job/PreCommit-YARN-Build/1939//console This message is automatically generated. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Karthik Kambatla Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.12.patch, yarn-353-12-wip.patch, YARN-353.13.patch, YARN-353.14.patch, YARN-353.15.patch, YARN-353.16.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13768721#comment-13768721 ] Karthik Kambatla commented on YARN-353: --- Thanks again for the detailed review, [~hitesh]. Sorry for the slow turnaround on this. {quote} {code} } catch (Exception e) { e.printStackTrace(); Assert.fail(ZKRMStateStore Session restore failed); } {code} {quote} bq. Don't think there is any need to catch the exception. The unit test will fail if the exception is not caught. Agree. The difference is in what the test reports shows this as - Failure or Error. The general rule I try to follow is - errors encountered in the specific code we are testing should be marked failures. I am not particular about this though, let me know if you would like me to make that change as well. The updated patch addresses some of these inconsistencies as well. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Karthik Kambatla Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.12.patch, yarn-353-12-wip.patch, YARN-353.13.patch, YARN-353.14.patch, YARN-353.15.patch, YARN-353.16.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13761573#comment-13761573 ] Hitesh Shah commented on YARN-353: -- hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/pom.xml - shouldn't there be a dependency on zookeeper even in the normal scope? {code} + if (childNodeName.startsWith(DELEGATION_TOKEN_SEQUENCE_NUMBER_PREFIX)) { +rmState.rmSecretManagerState.dtSequenceNumber = +Integer.parseInt(childNodeName.split(_)[1]); +continue; + } {code} - Could you clarify whether there can be multiple child nodes prefixed with DELEGATION_TOKEN_SEQUENCE_NUMBER_PREFIX in any possible state variation? {code} + // assert child node name is same as actual applicationId + assert appId.equals(appState.context.getApplicationId()); {code} - why the need for an assert? Should this check throw a runtime exception instead? (likewise for other assert checks ) {code} +} catch (Exception e) { + // currently throw all exceptions. May need to respond differently for HA + // based on whether we have lost the right to write to ZK + // TODO: Revisit this post YARN-149 + throw e; +} {code} - I believe its better to just remove such code and add it in with HA patches. {code} +/** + * Call exists() to leave a watch on the node denoted by path. + * Delete node if exists. To pass the existence information to the + * caller, call delete irrespective of whether node exists or not. + */ +if (zkClient.exists(path, true) == null) { + LOG.error(Trying to delete a path ( + path + + ) that doesn't exist.); +} else { + zkClient.delete(path, version); +} {code} - code does not match the comment ( with respect to passing of existence information ) {code} +} catch (Exception e) { + e.printStackTrace(); + Assert.fail(ZKRMStateStore Session restore failed); +} {code} - Don't think there is any need to catch the exception. The unit test will fail if the exception is not caught. If the exception stack trace in the unit test logs is not useful enough to understand the failure reason, it may be better to fix the code as needed. ( likewise in the other couple of places in the unit test where exceptions are being caught and handled with an assert.fail() {code} +Thread.sleep(800); {code} - the zk unit test has magic sleeps of 800 ms in some cases, 500 ms in others. What is the reason for these different numbers? Does the test helper need augmenting to remove this timing related dependency? General minor nits: - 80 chars line limit exceeded in multiple files. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Karthik Kambatla Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.12.patch, yarn-353-12-wip.patch, YARN-353.13.patch, YARN-353.14.patch, YARN-353.15.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13759408#comment-13759408 ] Karthik Kambatla commented on YARN-353: --- [~hitesh], when you get a chance, can you take a look at the updated patch. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Karthik Kambatla Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.12.patch, yarn-353-12-wip.patch, YARN-353.13.patch, YARN-353.14.patch, YARN-353.15.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13752755#comment-13752755 ] Hadoop QA commented on YARN-353: {color:green}+1 overall{color}. Here are the results of testing the latest attachment http://issues.apache.org/jira/secure/attachment/12600180/YARN-353.15.patch against trunk revision . {color:green}+1 @author{color}. The patch does not contain any @author tags. {color:green}+1 tests included{color}. The patch appears to include 3 new or modified test files. {color:green}+1 javac{color}. The applied patch does not increase the total number of javac compiler warnings. {color:green}+1 javadoc{color}. The javadoc tool did not generate any warning messages. {color:green}+1 eclipse:eclipse{color}. The patch built with eclipse:eclipse. {color:green}+1 findbugs{color}. The patch does not introduce any new Findbugs (version 1.3.9) warnings. {color:green}+1 release audit{color}. The applied patch does not increase the total number of release audit warnings. {color:green}+1 core tests{color}. The patch passed unit tests in hadoop-common-project/hadoop-common hadoop-yarn-project/hadoop-yarn/hadoop-yarn-api hadoop-yarn-project/hadoop-yarn/hadoop-yarn-common hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager. {color:green}+1 contrib tests{color}. The patch passed contrib unit tests. Test results: https://builds.apache.org/job/PreCommit-YARN-Build/1783//testReport/ Console output: https://builds.apache.org/job/PreCommit-YARN-Build/1783//console This message is automatically generated. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Karthik Kambatla Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.12.patch, yarn-353-12-wip.patch, YARN-353.13.patch, YARN-353.14.patch, YARN-353.15.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13750825#comment-13750825 ] Karthik Kambatla commented on YARN-353: --- Thanks for the detailed review, [~hitesh]. YARN-353.14.patch rebases against trunk and addresses most of the comments, but depends on HADOOP-9906. bq. FS State store uses fs.rm-state-store where as ZK uses zk.state-store - this is inconsistent. YARN-1056 fixes the FSRMStateStore prefix. Now, they are consistent. bq. Also, the variable names seem a bit inconsistent - should they be RM_ZK_STATE_STORE* as compared to ZK_RM_STATE_STORE* to match the actual property names? Though the property name itself has RM defined twice instead of just once. The variable names mimic the class names of the RMStateStore implementations. I believe this reads better. bq. LOG.debug statement not encapsulated within if isDebugEnabled() in RMStateStore.java. Fixed. bq. Could you add comments so that this piece of code is removed when HA handling work is done. Fixed and filed YARN-1099 for the same. bq. Why are all exceptions being caught instead of an explicit set? If there is a good reason for all exceptions, why not catch Throwable to capture everything? bq. The reason for why a connection to ZK failed is never logged. The only message logged is failed to connect on attempt X and that too only in debug level. Fixed. bq. It seems like throws Exception is being used in most places as compared to known types? Looked through and fixed where possible. bq. The tests for basic zk connect/CRUD probably belong in a separate file. Moved to a separate file. bq. where is this message being produced? is the RM code validating the format or ZK, and if ZK, should we be testing this in the first place? Assuming the test is to validate that we are using the configured value properly, how about a test for a diff perm from the default and checking that the zkNode has that permission set. Fixed through HADOOP-9906 and changes in this patch. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Karthik Kambatla Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.12.patch, yarn-353-12-wip.patch, YARN-353.13.patch, YARN-353.14.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13750830#comment-13750830 ] Hadoop QA commented on YARN-353: {color:red}-1 overall{color}. Here are the results of testing the latest attachment http://issues.apache.org/jira/secure/attachment/12600068/YARN-353.14.patch against trunk revision . {color:green}+1 @author{color}. The patch does not contain any @author tags. {color:green}+1 tests included{color}. The patch appears to include 3 new or modified test files. {color:red}-1 javac{color:red}. The patch appears to cause the build to fail. Console output: https://builds.apache.org/job/PreCommit-YARN-Build/1773//console This message is automatically generated. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Karthik Kambatla Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.12.patch, yarn-353-12-wip.patch, YARN-353.13.patch, YARN-353.14.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13745516#comment-13745516 ] Hitesh Shah commented on YARN-353: -- Sorry for the delay in the review. Been sidetracked by other work. Comments: FS State store uses fs.rm-state-store where as ZK uses zk.state-store - this is inconsistent. Also, the variable names seem a bit inconsistent - should they be RM_ZK_STATE_STORE* as compared to ZK_RM_STATE_STORE* to match the actual property names? Though the property name itself has RM defined twice instead of just once. Use of This must be supplied when using org.apache.hadoop.yarn.server.resourcemanager.recovery.ZKRMStateStore in yarn-default.xml - is the whole class name required? If yes, there are 1-2 properties which do not use the full class name. LOG.debug statement not encapsulated within if isDebugEnabled() in RMStateStore.java. {code} +} catch (Exception e) { + throw e; +} {code} - Could you add comments so that this piece of code is removed when HA handling work is done. {code} + try { +zkClient = getNewZooKeeper(); + } catch (Exception e) { +if (LOG.isDebugEnabled()) { + LOG.debug(Failed to connect to the ZooKeeper on attempt - + + (retries + 1)); +} + } {code} - Why are all exceptions being caught instead of an explicit set? If there is a good reason for all exceptions, why not catch Throwable to capture everything? - how is a failure in closing of zk connections meant to be handled in the createConnection function? - The reason for why a connection to ZK failed is never logged. The only message logged is failed to connect on attempt X and that too only in debug level. It seems like throws Exception is being used in most places as compared to known types? The tests for basic zk connect/CRUD probably belong in a separate file. {code} + Assert.assertTrue(e.getMessage().contains( +not of expected form scheme:id:perm)); {code} - where is this message being produced? is the RM code validating the format or ZK, and if ZK, should we be testing this in the first place? Assuming the test is to validate that we are using the configured value properly, how about a test for a diff perm from the default and checking that the zkNode has that permission set. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Karthik Kambatla Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.12.patch, yarn-353-12-wip.patch, YARN-353.13.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13736688#comment-13736688 ] Karthik Kambatla commented on YARN-353: --- Tested the latest patch on pseudo-dist cluster and ran into the same problem as YARN-1058. The patch behaves very similar to the FileSystemRMStateStore implementation. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.12.patch, yarn-353-12-wip.patch, YARN-353.13.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13736695#comment-13736695 ] Karthik Kambatla commented on YARN-353: --- Assigned to myself for easier tracking. The patch is primarily implemented by Bikas. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Karthik Kambatla Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.12.patch, yarn-353-12-wip.patch, YARN-353.13.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13737667#comment-13737667 ] Hadoop QA commented on YARN-353: {color:red}-1 overall{color}. Here are the results of testing the latest attachment http://issues.apache.org/jira/secure/attachment/12597420/YARN-353.13.patch against trunk revision . {color:green}+1 @author{color}. The patch does not contain any @author tags. {color:green}+1 tests included{color}. The patch appears to include 2 new or modified test files. {color:green}+1 javac{color}. The applied patch does not increase the total number of javac compiler warnings. {color:green}+1 javadoc{color}. The javadoc tool did not generate any warning messages. {color:green}+1 eclipse:eclipse{color}. The patch built with eclipse:eclipse. {color:red}-1 findbugs{color}. The patch appears to introduce 2 new Findbugs (version 1.3.9) warnings. {color:green}+1 release audit{color}. The applied patch does not increase the total number of release audit warnings. {color:green}+1 core tests{color}. The patch passed unit tests in hadoop-common-project/hadoop-common hadoop-yarn-project/hadoop-yarn/hadoop-yarn-api hadoop-yarn-project/hadoop-yarn/hadoop-yarn-common hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager. {color:green}+1 contrib tests{color}. The patch passed contrib unit tests. Test results: https://builds.apache.org/job/PreCommit-YARN-Build/1699//testReport/ Findbugs warnings: https://builds.apache.org/job/PreCommit-YARN-Build/1699//artifact/trunk/patchprocess/newPatchFindbugsWarningshadoop-common.html Console output: https://builds.apache.org/job/PreCommit-YARN-Build/1699//console This message is automatically generated. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Karthik Kambatla Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.12.patch, yarn-353-12-wip.patch, YARN-353.13.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13736035#comment-13736035 ] Jian He commented on YARN-353: -- Looks good overall, few nits {code} for (int retries = 0; retries numRetries zkClient == null; retries++) { zkClient = getNewZooKeeper(); } {code} If getNewZooKeeper throws an Exception, it will break out of the loop, instead of retry again ? is it necessary to put a try/catch block around it ? bq. We are calling createWithRetries on multiple roots 'zkRootNodePath' is in fact the common root for 'rmDTSecretManagerRoot' and 'rmAppRoot', or we can at least new ZooKeeper along with'znodeWorkingPath'? I missed one change in my earlier patch, change 'yarn.resourcemanager.zk.rm-state-store.address' to 'yarn.resourcemanager.zk.state-store.address', can you fix this, thanks. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.12.patch, yarn-353-12-wip.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13736196#comment-13736196 ] Hadoop QA commented on YARN-353: {color:red}-1 overall{color}. Here are the results of testing the latest attachment http://issues.apache.org/jira/secure/attachment/12597323/YARN-353.13.patch against trunk revision . {color:red}-1 patch{color}. Trunk compilation may be broken. Console output: https://builds.apache.org/job/PreCommit-YARN-Build/1691//console This message is automatically generated. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.12.patch, yarn-353-12-wip.patch, YARN-353.13.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13736207#comment-13736207 ] Hadoop QA commented on YARN-353: {color:red}-1 overall{color}. Here are the results of testing the latest attachment http://issues.apache.org/jira/secure/attachment/12597326/YARN-353.13.patch against trunk revision . {color:red}-1 patch{color}. Trunk compilation may be broken. Console output: https://builds.apache.org/job/PreCommit-YARN-Build/1692//console This message is automatically generated. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.12.patch, yarn-353-12-wip.patch, YARN-353.13.patch, YARN-353.13.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13734473#comment-13734473 ] Karthik Kambatla commented on YARN-353: --- bq. Identation still off in yarn-default.xml bq. hadoop-yarn-server-resourcemanager/pom.xml is still not fixed - contains the version info. LOG.debug() should be wrapped within if LOG.isDebugEnabled() {} bq. In abstract class ZKActionT, member variables should be final. bq. findbugs issue - zkSessionTimeout does not seem to be an issue but zkClient is something which is read/modified multiple times in various functions. Which functions are the ones that are/cannot be synchronized that access zkClient? bq. os.close(); - close calls should be in a finally block ( in multiple places ). bq. LOG.info(Error in storing + dtSequenceNumberPath); - change to LOG.info(Error in storing + dtSequenceNumberPath, e); bq. LOG.info(Created new connection for + this); -- logging 'this'? is there a toString() function? bq. How is a connection failure to zk handled? i.e. getNewZooKeeper() throws an exception. Does the RM fail/shutdown? Is the connection retried at a later point? bq. what if the first line throws an exception saying node exists but the other nodes are not created? Shouldn't each call be in its own try catch block? Or should the create function be changed to accept a parameter which when set causes the function to ignore node exists errors? bq. For deleteWithRetries, the return code of exists() could be checked if a delete is required or not. Fixed. bq. When creating a Zookeeper object, ZK apis support a base root path and all operations are done relative to the base root path? Any reason why we are not using that approach by initializing zk with zkRootNodePath ? We are calling createWithRetries on multiple roots bq. why the catch and re-throw? ( in multiple places ) The HA code might want to deal with those scenarios differently. Given we are on the verge of implement ZK-based HA, I think it is okay to leave these as they currently are. bq. validation only for non-null and not a valid format? While it is a good idea to check for a valid format, the current behavior is how RM etc. deal with host:port arguments bq. Not sure if a default value of !-value127.0.0.1:2181/value- should be mentioned in yarn-default.xml. Changed the comment to be host:port. Leaving it there makes it easier for users. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.12.patch, yarn-353-12-wip.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13734570#comment-13734570 ] Hadoop QA commented on YARN-353: {color:red}-1 overall{color}. Here are the results of testing the latest attachment http://issues.apache.org/jira/secure/attachment/12597034/YARN-353.12.patch against trunk revision . {color:green}+1 @author{color}. The patch does not contain any @author tags. {color:green}+1 tests included{color}. The patch appears to include 2 new or modified test files. {color:green}+1 javac{color}. The applied patch does not increase the total number of javac compiler warnings. {color:green}+1 javadoc{color}. The javadoc tool did not generate any warning messages. {color:green}+1 eclipse:eclipse{color}. The patch built with eclipse:eclipse. {color:red}-1 findbugs{color}. The patch appears to introduce 2 new Findbugs (version 1.3.9) warnings. {color:green}+1 release audit{color}. The applied patch does not increase the total number of release audit warnings. {color:green}+1 core tests{color}. The patch passed unit tests in hadoop-common-project/hadoop-common hadoop-yarn-project/hadoop-yarn/hadoop-yarn-api hadoop-yarn-project/hadoop-yarn/hadoop-yarn-common hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager. {color:green}+1 contrib tests{color}. The patch passed contrib unit tests. Test results: https://builds.apache.org/job/PreCommit-YARN-Build/1681//testReport/ Findbugs warnings: https://builds.apache.org/job/PreCommit-YARN-Build/1681//artifact/trunk/patchprocess/newPatchFindbugsWarningshadoop-common.html Console output: https://builds.apache.org/job/PreCommit-YARN-Build/1681//console This message is automatically generated. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.12.patch, yarn-353-12-wip.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13735088#comment-13735088 ] Karthik Kambatla commented on YARN-353: --- Not quite sure why we see the findbugs warning, it is in files that I haven't touched. Filed HDFS-5082 to handle remove zookeeper version info from hadoop-hdfs/pom.xml. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.12.patch, yarn-353-12-wip.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13733576#comment-13733576 ] Hitesh Shah commented on YARN-353: -- bq. For deleteWithRetries, the return code of exists() could be checked if a delete is required or not. this depends on whether RM wants to know the delete operation succeeds or not. I am not sure I understand. If the RM is trying to delete something and the node does not exist, is there a situation where the RM wants to know that the node didn't exist and fail if a non-existent node was tried to be deleted? Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13733684#comment-13733684 ] Jian He commented on YARN-353: -- bq. I am not sure I understand. If the RM is trying to delete something and the node does not exist, is there a situation where the RM wants to know that the node didn't exist and fail if a non-existent node was tried to be deleted? Agreed. We should specifically check if the node exists or not. Otherwise the ZK delete() API will throw an exception if node doesn't exist which we don't want. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13733760#comment-13733760 ] Karthik Kambatla commented on YARN-353: --- Looking into this now. Will hopefully have an update (patch + replies) sometime today. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13732868#comment-13732868 ] Hitesh Shah commented on YARN-353: -- Comments: Identation still off in yarn-default.xml Not sure if a default value of !--value127.0.0.1:2181/value-- should be mentioned in yarn-default.xml. hadoop-yarn-server-resourcemanager/pom.xml is still not fixed - contains the version info. LOG.debug() should be wrapped within if LOG.isDebugEnabled() {} In abstract class ZKActionT, member variables should be final. findbugs issue - zkSessionTimeout does not seem to be an issue but zkClient is something which is read/modified multiple times in various functions. Which functions are the ones that are/cannot be synchronized that access zkClient? {code} +if (zkHostPort == null) { + throw new YarnRuntimeException( +No server address specified for zookeeper state store for Resource + + Manager recovery. ZK_RM_STATE_STORE_ADDRESS is not configured.); {code} - validation only for non-null and not a valid format? - ZK_RM_STATE_STORE_ADDRESS as a string? A question on general usage of zk: why is everything being stored at the top level of the tree and not a heirarchical structure i.e. attempts of a particular application stored under that application's dir? {code} +} catch (Exception e) { + throw e; +} {code} - why the catch and re-throw? ( in multiple places ) os.close(); - close calls should be in a finally block ( in multiple places ). LOG.info(Error in storing + dtSequenceNumberPath); - change to LOG.info(Error in storing + dtSequenceNumberPath, e); {code} + String getNodePath(String root, String nodeName) { +return (root + / + nodeName); + } {code} - does ZK have a variable to define the node path separator which we can use instead of a magic string? LOG.info(Created new connection for + this); -- logging 'this'? is there a toString() function? How is a connection failure to zk handled? i.e. getNewZooKeeper() throws an exception. Does the RM fail/shutdown? Is the connection retried at a later point? When creating a Zookeeper object, ZK apis support a base root path and all operations are done relative to the base root path? Any reason why we are not using that approach by initializing zk with zkRootNodePath ? {code} +try { + createWithRetries(znodeWorkingPath, null, zkAcl, CreateMode.PERSISTENT); + createWithRetries(zkRootNodePath, null, zkAcl, CreateMode.PERSISTENT); + createWithRetries(rmDTSecretManagerRoot, null, zkAcl, +CreateMode.PERSISTENT); + createWithRetries(rmAppRoot, null, zkAcl, CreateMode.PERSISTENT); +} catch (KeeperException ke) { + if (ke.code() != Code.NODEEXISTS) { +throw ke; + } +} {code} - what if the first line throws an exception saying node exists but the other nodes are not created? Shouldn't each call be in its own try catch block? Or should the create function be changed to accept a parameter which when set causes the function to ignore node exists errors? For deleteWithRetries, the return code of exists() could be checked if a delete is required or not. Still need to look at the unit tests. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13733127#comment-13733127 ] Jian He commented on YARN-353: -- Go ahead, thanks for that, Karthik ! Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13733164#comment-13733164 ] Jian He commented on YARN-353: -- bq. why the catch and re-throw? The intention was leaving there for future exception handling(HA etc). could be removed. bq. why is everything being stored at the top level of the tree and not a heirarchical structure i.e. attempts of a particular application stored under that application's dir? ZK doesn't support directory bq. How is a connection failure to zk handled? i.e. getNewZooKeeper() throws an exception. Does the RM fail/shutdown? Is the connection retried at a later point? If the connection is lost after RM already connected with ZK, a Disconnected event will be sent, zkClient sets to null, all operations(create/delete etc.) will wait for zkClient to be set again (verified this in cluster by stopping ZK during the connection). But if getNewZooKeeper() itself throws an exception, for now RM will fail and will not retry bq.When creating a Zookeeper object, ZK apis support a base root path and all operations are done relative to the base root path? Any reason why we are not using that approach by initializing zk with zkRootNodePath ? didn't know this. bq.For deleteWithRetries, the return code of exists() could be checked if a delete is required or not. this depends on whether RM wants to know the delete operation succeeds or not. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13731480#comment-13731480 ] Karthik Kambatla commented on YARN-353: --- YARN-353.11.patch is the patch with findbugs exclusions. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13731528#comment-13731528 ] Hadoop QA commented on YARN-353: {color:green}+1 overall{color}. Here are the results of testing the latest attachment http://issues.apache.org/jira/secure/attachment/12596465/YARN-353.11.patch against trunk revision . {color:green}+1 @author{color}. The patch does not contain any @author tags. {color:green}+1 tests included{color}. The patch appears to include 2 new or modified test files. {color:green}+1 javac{color}. The applied patch does not increase the total number of javac compiler warnings. {color:green}+1 javadoc{color}. The javadoc tool did not generate any warning messages. {color:green}+1 eclipse:eclipse{color}. The patch built with eclipse:eclipse. {color:green}+1 findbugs{color}. The patch does not introduce any new Findbugs (version 1.3.9) warnings. {color:green}+1 release audit{color}. The applied patch does not increase the total number of release audit warnings. {color:green}+1 core tests{color}. The patch passed unit tests in hadoop-common-project/hadoop-common hadoop-yarn-project/hadoop-yarn/hadoop-yarn-api hadoop-yarn-project/hadoop-yarn/hadoop-yarn-common hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager. {color:green}+1 contrib tests{color}. The patch passed contrib unit tests. Test results: https://builds.apache.org/job/PreCommit-YARN-Build/1665//testReport/ Console output: https://builds.apache.org/job/PreCommit-YARN-Build/1665//console This message is automatically generated. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.10.patch, YARN-353.11.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13729051#comment-13729051 ] Jian He commented on YARN-353: -- The findbug -1 might be related to the test case is non-synchronously referencing 'zkClient' and 'zkSessionTimeout', which may not be an issue. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.10.patch, YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13727391#comment-13727391 ] Hitesh Shah commented on YARN-353: -- [~jianhe] Could you please look at the findbug warnings raised by the patch? Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13727174#comment-13727174 ] Hadoop QA commented on YARN-353: {color:red}-1 overall{color}. Here are the results of testing the latest attachment http://issues.apache.org/jira/secure/attachment/12595519/YARN-353.9.patch against trunk revision . {color:green}+1 @author{color}. The patch does not contain any @author tags. {color:green}+1 tests included{color}. The patch appears to include 2 new or modified test files. {color:green}+1 javac{color}. The applied patch does not increase the total number of javac compiler warnings. {color:green}+1 javadoc{color}. The javadoc tool did not generate any warning messages. {color:green}+1 eclipse:eclipse{color}. The patch built with eclipse:eclipse. {color:red}-1 findbugs{color}. The patch appears to introduce 2 new Findbugs (version 1.3.9) warnings. {color:green}+1 release audit{color}. The applied patch does not increase the total number of release audit warnings. {color:green}+1 core tests{color}. The patch passed unit tests in hadoop-common-project/hadoop-common hadoop-yarn-project/hadoop-yarn/hadoop-yarn-api hadoop-yarn-project/hadoop-yarn/hadoop-yarn-common hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager. {color:green}+1 contrib tests{color}. The patch passed contrib unit tests. Test results: https://builds.apache.org/job/PreCommit-YARN-Build/1644//testReport/ Findbugs warnings: https://builds.apache.org/job/PreCommit-YARN-Build/1644//artifact/trunk/patchprocess/newPatchFindbugsWarningshadoop-yarn-server-resourcemanager.html Console output: https://builds.apache.org/job/PreCommit-YARN-Build/1644//console This message is automatically generated. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch, YARN-353.9.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13724585#comment-13724585 ] Sandy Ryza commented on YARN-353: - The patch looks good to me, other than the findbugs warning and {code} +LOG + .info(Removing RMDelegationToken_ + rmDTIdentifier.getSequenceNumber()); {code} which should be broken up to have LOG and .info on the same line. Have any manual steps been performed to verify that this works on a cluster? Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13712523#comment-13712523 ] Bikas Saha commented on YARN-353: - bq. ZKRMStateStore#getNewZooKeeper need not be synchronized bq. fixed The code is derived from ActiveStandyLeaderElector code in hadoop common. It was synchronized there for a race condition that showed up in testing. I would like to keep the synchronization as it was in the original patch. bq. the patch still seems to have NUM_RETRIES Why should NUM_RETRIES not be there? Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13712533#comment-13712533 ] Karthik Kambatla commented on YARN-353: --- bq. Make the ZKRMStateStore#NUM_RETRIES configurable with default set to 3. bq. fixed bq. Why should NUM_RETRIES not be there? Was just noting that: the latest patch has the non-configurable NUM_RETRIES, it should exist but be configurable. If it is configurable, we should probably change the name of the variable. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13712793#comment-13712793 ] Karthik Kambatla commented on YARN-353: --- Looks good. +1 pending Jenkins. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13712902#comment-13712902 ] Hadoop QA commented on YARN-353: {color:red}-1 overall{color}. Here are the results of testing the latest attachment http://issues.apache.org/jira/secure/attachment/12593045/YARN-353.8.patch against trunk revision . {color:green}+1 @author{color}. The patch does not contain any @author tags. {color:green}+1 tests included{color}. The patch appears to include 2 new or modified test files. {color:green}+1 javac{color}. The applied patch does not increase the total number of javac compiler warnings. {color:green}+1 javadoc{color}. The javadoc tool did not generate any warning messages. {color:green}+1 eclipse:eclipse{color}. The patch built with eclipse:eclipse. {color:red}-1 findbugs{color}. The patch appears to introduce 3 new Findbugs (version 1.3.9) warnings. {color:green}+1 release audit{color}. The applied patch does not increase the total number of release audit warnings. {color:green}+1 core tests{color}. The patch passed unit tests in hadoop-common-project/hadoop-common hadoop-yarn-project/hadoop-yarn/hadoop-yarn-api hadoop-yarn-project/hadoop-yarn/hadoop-yarn-common hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager. {color:green}+1 contrib tests{color}. The patch passed contrib unit tests. Test results: https://builds.apache.org/job/PreCommit-YARN-Build/1520//testReport/ Findbugs warnings: https://builds.apache.org/job/PreCommit-YARN-Build/1520//artifact/trunk/patchprocess/newPatchFindbugsWarningshadoop-yarn-server-resourcemanager.html Console output: https://builds.apache.org/job/PreCommit-YARN-Build/1520//console This message is automatically generated. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13712931#comment-13712931 ] Karthik Kambatla commented on YARN-353: --- For the findbugs warning around NUM_RETRIES, we should probably make it non-static numRetries. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch, YARN-353.8.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13708734#comment-13708734 ] Jian He commented on YARN-353: -- Thanks for the review, Karthik bq.YarnConfiguration: how about creating a common prefix for all of zk-state-store related parameters? bq.Make the ZKRMStateStore#NUM_RETRIES configurable with default set to 3. bq.ZKRMStateStore#getNewZooKeeper need not be synchronized fixed bq.Might be cleaner to move zkDoWithRetries to ZkAction we can implement no-retry functionalities with ZkAction if separate zkDoWithRetries out of ZkAction. same reason for 6 New patch also added test case for ZKClient disconnect and reconnect logic. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13708796#comment-13708796 ] Hadoop QA commented on YARN-353: {color:red}-1 overall{color}. Here are the results of testing the latest attachment http://issues.apache.org/jira/secure/attachment/12592363/YARN-353.6.patch against trunk revision . {color:green}+1 @author{color}. The patch does not contain any @author tags. {color:green}+1 tests included{color}. The patch appears to include 2 new or modified test files. {color:green}+1 javac{color}. The applied patch does not increase the total number of javac compiler warnings. {color:green}+1 javadoc{color}. The javadoc tool did not generate any warning messages. {color:green}+1 eclipse:eclipse{color}. The patch built with eclipse:eclipse. {color:red}-1 findbugs{color}. The patch appears to introduce 3 new Findbugs (version 1.3.9) warnings. {color:green}+1 release audit{color}. The applied patch does not increase the total number of release audit warnings. {color:green}+1 core tests{color}. The patch passed unit tests in hadoop-common-project/hadoop-common hadoop-yarn-project/hadoop-yarn/hadoop-yarn-api hadoop-yarn-project/hadoop-yarn/hadoop-yarn-common hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager. {color:green}+1 contrib tests{color}. The patch passed contrib unit tests. Test results: https://builds.apache.org/job/PreCommit-YARN-Build/1482//testReport/ Findbugs warnings: https://builds.apache.org/job/PreCommit-YARN-Build/1482//artifact/trunk/patchprocess/newPatchFindbugsWarningshadoop-yarn-server-resourcemanager.html Console output: https://builds.apache.org/job/PreCommit-YARN-Build/1482//console This message is automatically generated. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13709060#comment-13709060 ] Hadoop QA commented on YARN-353: {color:red}-1 overall{color}. Here are the results of testing the latest attachment http://issues.apache.org/jira/secure/attachment/12592413/YARN-353.7.patch against trunk revision . {color:green}+1 @author{color}. The patch does not contain any @author tags. {color:green}+1 tests included{color}. The patch appears to include 2 new or modified test files. {color:green}+1 javac{color}. The applied patch does not increase the total number of javac compiler warnings. {color:green}+1 javadoc{color}. The javadoc tool did not generate any warning messages. {color:green}+1 eclipse:eclipse{color}. The patch built with eclipse:eclipse. {color:red}-1 findbugs{color}. The patch appears to introduce 1 new Findbugs (version 1.3.9) warnings. {color:green}+1 release audit{color}. The applied patch does not increase the total number of release audit warnings. {color:red}-1 core tests{color}. The patch failed these unit tests in hadoop-common-project/hadoop-common hadoop-yarn-project/hadoop-yarn/hadoop-yarn-api hadoop-yarn-project/hadoop-yarn/hadoop-yarn-common hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager: org.apache.hadoop.yarn.server.resourcemanager.recovery.TestRMStateStore {color:green}+1 contrib tests{color}. The patch passed contrib unit tests. Test results: https://builds.apache.org/job/PreCommit-YARN-Build/1484//testReport/ Findbugs warnings: https://builds.apache.org/job/PreCommit-YARN-Build/1484//artifact/trunk/patchprocess/newPatchFindbugsWarningshadoop-yarn-server-resourcemanager.html Console output: https://builds.apache.org/job/PreCommit-YARN-Build/1484//console This message is automatically generated. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch, YARN-353.6.patch, YARN-353.7.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13703608#comment-13703608 ] Hadoop QA commented on YARN-353: {color:green}+1 overall{color}. Here are the results of testing the latest attachment http://issues.apache.org/jira/secure/attachment/12591477/YARN-353.5.patch against trunk revision . {color:green}+1 @author{color}. The patch does not contain any @author tags. {color:green}+1 tests included{color}. The patch appears to include 1 new or modified test files. {color:green}+1 javac{color}. The applied patch does not increase the total number of javac compiler warnings. {color:green}+1 javadoc{color}. The javadoc tool did not generate any warning messages. {color:green}+1 eclipse:eclipse{color}. The patch built with eclipse:eclipse. {color:green}+1 findbugs{color}. The patch does not introduce any new Findbugs (version 1.3.9) warnings. {color:green}+1 release audit{color}. The applied patch does not increase the total number of release audit warnings. {color:green}+1 core tests{color}. The patch passed unit tests in hadoop-yarn-project/hadoop-yarn/hadoop-yarn-api hadoop-yarn-project/hadoop-yarn/hadoop-yarn-common hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager. {color:green}+1 contrib tests{color}. The patch passed contrib unit tests. Test results: https://builds.apache.org/job/PreCommit-YARN-Build/1439//testReport/ Console output: https://builds.apache.org/job/PreCommit-YARN-Build/1439//console This message is automatically generated. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch, YARN-353.5.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13700487#comment-13700487 ] Bikas Saha commented on YARN-353: - I really don't know how to explain the downside of having a bogus default ZK address to which the RM will try to connect. This should be obvious. Either that or I am seeing things differently from you. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13700490#comment-13700490 ] Devaraj K commented on YARN-353: If we see, {code:xml} public static final String RM_RESOURCE_TRACKER_ADDRESS = RM_PREFIX + resource-tracker.address; public static final int DEFAULT_RM_RESOURCE_TRACKER_PORT = 8031; public static final String DEFAULT_RM_RESOURCE_TRACKER_ADDRESS = 0.0.0.0: + DEFAULT_RM_RESOURCE_TRACKER_PORT; {code} Here also we should configure the RM address if the RM is running in other machine than NM, otherwise it will be fine. Similarly we can provide the default value for zk address and will be useful if it is single node cluster or test environment. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13700371#comment-13700371 ] Bikas Saha commented on YARN-353: - I dont think it makes sense to have default value for this. ZK location is not something we control and we cannot assume it to be running on some default location. The commented value in the default.xml file is just for a syntax example. {code} + public static final String DEFAULT_ZK_RM_STATE_STORE_ADDRESS = + 127.0.0.1:2181; {code} Wherever we are doing multiple operations, we should probably use the ZK multi API's to guarantee atomic operations. {code} ++ latestSequenceNumber); +try { + if (dtSequenceNumberPath != null) { +deleteWithRetries(dtSequenceNumberPath, 0); + } + createWithRetries(latestSequenceNumberPath, null, zkAcl, +CreateMode.PERSISTENT); +} catch (Exception e) { + LOG.info(Error in storing + dtSequenceNumberPath); + throw e; +} +dtSequenceNumberPath = latestSequenceNumberPath; {code} Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13700395#comment-13700395 ] Jian He commented on YARN-353: -- bq. I dont think it makes sense to have default value for this. ZK location is not something we control and we cannot assume it to be running on some default location. Yes, we can not assume which location ZK is ruining on, but I think the result would be the same if we provide a default or leave it empty, botch cases should raise connect exception or something, which leads the user to config the true address. One bonus doing such might make user easier in test mode where ZK is running on its defaults, your opinion? Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13700398#comment-13700398 ] Bikas Saha commented on YARN-353: - No. It must be required for the user to specify this. We cannot assume some random address if the user has not specified a value. The code should throw an exception if this is not specified. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13700422#comment-13700422 ] Jian He commented on YARN-353: -- Any downside of doing that ? Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13700440#comment-13700440 ] Bikas Saha commented on YARN-353: - Downside of doing what? Throwing clear exception will alert the user that the address is not configured and so the RM will not start. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13700465#comment-13700465 ] Jian He commented on YARN-353: -- sorry, I meant downside of giving a default ZK address. yeah, throwing an exception would be clear. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13698750#comment-13698750 ] Devaraj K commented on YARN-353: The latest patch looks good to me except one nit. {code:xml} + public static final String DEFAULT_ZK_RM_STATE_STORE_PARENT_PATH = rmstore; {code} Here this path should start with '/', otherwise zkclient will throw IllegalArgumentException. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13699305#comment-13699305 ] Hadoop QA commented on YARN-353: {color:green}+1 overall{color}. Here are the results of testing the latest attachment http://issues.apache.org/jira/secure/attachment/12590691/YARN-353.4.patch against trunk revision . {color:green}+1 @author{color}. The patch does not contain any @author tags. {color:green}+1 tests included{color}. The patch appears to include 1 new or modified test files. {color:green}+1 javac{color}. The applied patch does not increase the total number of javac compiler warnings. {color:green}+1 javadoc{color}. The javadoc tool did not generate any warning messages. {color:green}+1 eclipse:eclipse{color}. The patch built with eclipse:eclipse. {color:green}+1 findbugs{color}. The patch does not introduce any new Findbugs (version 1.3.9) warnings. {color:green}+1 release audit{color}. The applied patch does not increase the total number of release audit warnings. {color:green}+1 core tests{color}. The patch passed unit tests in hadoop-yarn-project/hadoop-yarn/hadoop-yarn-api hadoop-yarn-project/hadoop-yarn/hadoop-yarn-common hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager. {color:green}+1 contrib tests{color}. The patch passed contrib unit tests. Test results: https://builds.apache.org/job/PreCommit-YARN-Build/1420//testReport/ Console output: https://builds.apache.org/job/PreCommit-YARN-Build/1420//console This message is automatically generated. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch, YARN-353.4.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13697629#comment-13697629 ] Devaraj K commented on YARN-353: The patch overall looks good, here are my observations on the patch. 1. {code:xml} + property +descriptionACL's to be used for ZooKeeper znodes. +This may be supplied when using +org.apache.hadoop.yarn.server.resourcemanager.recovery.ZKRMStateStore +as the value for yarn.resourcemanager.store.class/description +nameyarn.resourcemanager.zk.rm-state-store.timeout.ms/name +!--valueworld:anyone:rwcda/value-- + /property {code} Here configuration name should be yarn.resourcemanager.zk.rm-state-store.acl. 2. {code:xml} + // protected to mock for testing + protected synchronized ZooKeeper getNewZooKeeper() throws Exception { {code} Can we also annotate with @VisibleForTesting for this method? 3. {code:xml} + /** HostPort of ZK server for ZKRMStateStore */ +descriptionHostPort of the ZooKeeper server when using {code} These two places can we use Host:Port instead of HostPort for comment/description. 4. {code:xml} +zkHostPort = conf.get(YarnConfiguration.ZK_RM_STATE_STORE_ADDRESS); {code} Can we use the default value for this config with this as present for other props, {code:xml} +!--value127.0.0.1:2181/value-- {code} 5. {code:xml} + public static final String DEFAULT_ZK_RM_STATE_STORE_PARENT_PATH = ; {code} Can we use the default value for this config with this instead of having empty, {code:xml} +!--value/rmstore/value-- {code} Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.1.patch, YARN-353.2.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13698099#comment-13698099 ] Hadoop QA commented on YARN-353: {color:green}+1 overall{color}. Here are the results of testing the latest attachment http://issues.apache.org/jira/secure/attachment/12590474/YARN-353.3.patch against trunk revision . {color:green}+1 @author{color}. The patch does not contain any @author tags. {color:green}+1 tests included{color}. The patch appears to include 1 new or modified test files. {color:green}+1 javac{color}. The applied patch does not increase the total number of javac compiler warnings. {color:green}+1 javadoc{color}. The javadoc tool did not generate any warning messages. {color:green}+1 eclipse:eclipse{color}. The patch built with eclipse:eclipse. {color:green}+1 findbugs{color}. The patch does not introduce any new Findbugs (version 1.3.9) warnings. {color:green}+1 release audit{color}. The applied patch does not increase the total number of release audit warnings. {color:green}+1 core tests{color}. The patch passed unit tests in hadoop-yarn-project/hadoop-yarn/hadoop-yarn-api hadoop-yarn-project/hadoop-yarn/hadoop-yarn-common hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager. {color:green}+1 contrib tests{color}. The patch passed contrib unit tests. Test results: https://builds.apache.org/job/PreCommit-YARN-Build/1416//testReport/ Console output: https://builds.apache.org/job/PreCommit-YARN-Build/1416//console This message is automatically generated. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.1.patch, YARN-353.2.patch, YARN-353.3.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13697224#comment-13697224 ] Jian He commented on YARN-353: -- I'm taking this over Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.1.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13697417#comment-13697417 ] Hadoop QA commented on YARN-353: {color:red}-1 overall{color}. Here are the results of testing the latest attachment http://issues.apache.org/jira/secure/attachment/12590350/YARN-353.2.patch against trunk revision . {color:green}+1 @author{color}. The patch does not contain any @author tags. {color:green}+1 tests included{color}. The patch appears to include 1 new or modified test files. {color:green}+1 javac{color}. The applied patch does not increase the total number of javac compiler warnings. {color:green}+1 javadoc{color}. The javadoc tool did not generate any warning messages. {color:green}+1 eclipse:eclipse{color}. The patch built with eclipse:eclipse. {color:red}-1 findbugs{color}. The patch appears to introduce 3 new Findbugs (version 1.3.9) warnings. {color:green}+1 release audit{color}. The applied patch does not increase the total number of release audit warnings. {color:green}+1 core tests{color}. The patch passed unit tests in hadoop-yarn-project/hadoop-yarn/hadoop-yarn-api hadoop-yarn-project/hadoop-yarn/hadoop-yarn-common hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager. {color:green}+1 contrib tests{color}. The patch passed contrib unit tests. Test results: https://builds.apache.org/job/PreCommit-YARN-Build/1414//testReport/ Findbugs warnings: https://builds.apache.org/job/PreCommit-YARN-Build/1414//artifact/trunk/patchprocess/newPatchFindbugsWarningshadoop-yarn-server-resourcemanager.html Console output: https://builds.apache.org/job/PreCommit-YARN-Build/1414//console This message is automatically generated. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.1.patch, YARN-353.2.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13695656#comment-13695656 ] Karthik Kambatla commented on YARN-353: --- Thanks Bikas. Mostly looks good. Can you address the findbugs and rebase against trunk. Will post a detailed review (couple of nits) on the updated patch. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.1.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (YARN-353) Add Zookeeper-based store implementation for RMStateStore
[ https://issues.apache.org/jira/browse/YARN-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13561929#comment-13561929 ] Hadoop QA commented on YARN-353: {color:red}-1 overall{color}. Here are the results of testing the latest attachment http://issues.apache.org/jira/secure/attachment/12566356/YARN-353.1.patch against trunk revision . {color:green}+1 @author{color}. The patch does not contain any @author tags. {color:green}+1 tests included{color}. The patch appears to include 1 new or modified test files. {color:green}+1 javac{color}. The applied patch does not increase the total number of javac compiler warnings. {color:green}+1 javadoc{color}. The javadoc tool did not generate any warning messages. {color:green}+1 eclipse:eclipse{color}. The patch built with eclipse:eclipse. {color:red}-1 findbugs{color}. The patch appears to introduce 2 new Findbugs (version 1.3.9) warnings. {color:green}+1 release audit{color}. The applied patch does not increase the total number of release audit warnings. {color:green}+1 core tests{color}. The patch passed unit tests in hadoop-yarn-project/hadoop-yarn/hadoop-yarn-common hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager. {color:green}+1 contrib tests{color}. The patch passed contrib unit tests. Test results: https://builds.apache.org/job/PreCommit-YARN-Build/365//testReport/ Findbugs warnings: https://builds.apache.org/job/PreCommit-YARN-Build/365//artifact/trunk/patchprocess/newPatchFindbugsWarningshadoop-yarn-server-resourcemanager.html Console output: https://builds.apache.org/job/PreCommit-YARN-Build/365//console This message is automatically generated. Add Zookeeper-based store implementation for RMStateStore - Key: YARN-353 URL: https://issues.apache.org/jira/browse/YARN-353 Project: Hadoop YARN Issue Type: Sub-task Components: resourcemanager Reporter: Hitesh Shah Assignee: Bikas Saha Attachments: YARN-353.1.patch Add store that write RM state data to ZK -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira