[jira] [Commented] (HADOOP-17208) LoadBalanceKMSClientProvider#deleteKey should invalidateCache via all KMSClientProvider instances
[ https://issues.apache.org/jira/browse/HADOOP-17208?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17212819#comment-17212819 ] Xiaoyu Yao commented on HADOOP-17208: - I agree. With HADOOP-17304, this will not be no need to expose additional INVALIDATE_CACHE ACL for DELETE ops. The previous failed test can be used to validate this > LoadBalanceKMSClientProvider#deleteKey should invalidateCache via all > KMSClientProvider instances > - > > Key: HADOOP-17208 > URL: https://issues.apache.org/jira/browse/HADOOP-17208 > Project: Hadoop Common > Issue Type: Improvement >Affects Versions: 2.8.4 >Reporter: Xiaoyu Yao >Assignee: Xiaoyu Yao >Priority: Major > Labels: pull-request-available > Fix For: 3.4.0 > > Time Spent: 2h 20m > Remaining Estimate: 0h > > Without invalidateCache, the deleted key may still exists in the servers' key > cache (CachingKeyProvider in KMSWebApp.java) where the delete key was not > hit. Client may still be able to access encrypted files by specifying to > connect to KMS instances with a cached version of the deleted key before the > cache entry (10 min by default) expired. -- This message was sent by Atlassian Jira (v8.3.4#803005) - To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org
[jira] [Commented] (HADOOP-17208) LoadBalanceKMSClientProvider#deleteKey should invalidateCache via all KMSClientProvider instances
[ https://issues.apache.org/jira/browse/HADOOP-17208?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17212811#comment-17212811 ] Xiaoqiao He commented on HADOOP-17208: -- Thanks [~xyao] for your comments, I am concerned if this is incompatible improvement. After changes, we expose INVALIDATE_CACHE acl to end user which does not need to care about before. Please correct me if something I missed. Thanks. > LoadBalanceKMSClientProvider#deleteKey should invalidateCache via all > KMSClientProvider instances > - > > Key: HADOOP-17208 > URL: https://issues.apache.org/jira/browse/HADOOP-17208 > Project: Hadoop Common > Issue Type: Improvement >Affects Versions: 2.8.4 >Reporter: Xiaoyu Yao >Assignee: Xiaoyu Yao >Priority: Major > Labels: pull-request-available > Fix For: 3.4.0 > > Time Spent: 2h 20m > Remaining Estimate: 0h > > Without invalidateCache, the deleted key may still exists in the servers' key > cache (CachingKeyProvider in KMSWebApp.java) where the delete key was not > hit. Client may still be able to access encrypted files by specifying to > connect to KMS instances with a cached version of the deleted key before the > cache entry (10 min by default) expired. -- This message was sent by Atlassian Jira (v8.3.4#803005) - To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org
[jira] [Commented] (HADOOP-17208) LoadBalanceKMSClientProvider#deleteKey should invalidateCache via all KMSClientProvider instances
[ https://issues.apache.org/jira/browse/HADOOP-17208?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17212623#comment-17212623 ] Xiaoyu Yao commented on HADOOP-17208: - Good catch [~ayushtkn]. And thanks [~hexiaoqiao] for looking into this. The original design of the INVALIDATE_CACHE op is tied to ROLLOVER ACL. The test itself can be fixed by allowing "DELETE" user to have ROLLOVER just like SET_KEY_MATERIAL does. conf.set(KMSACLs.Type.ROLLOVER.getAclConfigKey(), KMSACLs.Type.ROLLOVER.toString() + ",SET_KEY_MATERIAL,DELETE"); It would be much clean if we can have a separate INVALIDATE_CACHE ACL type to differentiate INVALIDATE_CACHE other than ROLLOVER itself like SET_KEY_MATERIAL and DELETE. > LoadBalanceKMSClientProvider#deleteKey should invalidateCache via all > KMSClientProvider instances > - > > Key: HADOOP-17208 > URL: https://issues.apache.org/jira/browse/HADOOP-17208 > Project: Hadoop Common > Issue Type: Improvement >Affects Versions: 2.8.4 >Reporter: Xiaoyu Yao >Assignee: Xiaoyu Yao >Priority: Major > Labels: pull-request-available > Fix For: 3.4.0 > > Time Spent: 2h 20m > Remaining Estimate: 0h > > Without invalidateCache, the deleted key may still exists in the servers' key > cache (CachingKeyProvider in KMSWebApp.java) where the delete key was not > hit. Client may still be able to access encrypted files by specifying to > connect to KMS instances with a cached version of the deleted key before the > cache entry (10 min by default) expired. -- This message was sent by Atlassian Jira (v8.3.4#803005) - To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org
[jira] [Commented] (HADOOP-17208) LoadBalanceKMSClientProvider#deleteKey should invalidateCache via all KMSClientProvider instances
[ https://issues.apache.org/jira/browse/HADOOP-17208?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17212279#comment-17212279 ] Xiaoqiao He commented on HADOOP-17208: -- [~ayushtkn] Thanks for your comments. Failed unit test {{TestKMS.testACLs}} is related to change. The root cause is that not assign {{KMSACLs.Type.ROLLOVER}} privilege to user `DELETE` after add invalidate cache operation when delete key. It seems that this is not an incompatible improvement? ACL entry has changed for delete key after this changes. cc [~xyao] would you like to have another checks? Thanks. > LoadBalanceKMSClientProvider#deleteKey should invalidateCache via all > KMSClientProvider instances > - > > Key: HADOOP-17208 > URL: https://issues.apache.org/jira/browse/HADOOP-17208 > Project: Hadoop Common > Issue Type: Improvement >Affects Versions: 2.8.4 >Reporter: Xiaoyu Yao >Assignee: Xiaoyu Yao >Priority: Major > Labels: pull-request-available > Fix For: 3.4.0 > > Time Spent: 2h 20m > Remaining Estimate: 0h > > Without invalidateCache, the deleted key may still exists in the servers' key > cache (CachingKeyProvider in KMSWebApp.java) where the delete key was not > hit. Client may still be able to access encrypted files by specifying to > connect to KMS instances with a cached version of the deleted key before the > cache entry (10 min by default) expired. -- This message was sent by Atlassian Jira (v8.3.4#803005) - To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org
[jira] [Commented] (HADOOP-17208) LoadBalanceKMSClientProvider#deleteKey should invalidateCache via all KMSClientProvider instances
[ https://issues.apache.org/jira/browse/HADOOP-17208?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17211647#comment-17211647 ] Ayush Saxena commented on HADOOP-17208: --- [~xyao]/[~hexiaoqiao] This seems to have broken {{TestKMS.testACLs}} https://ci-hadoop.apache.org/job/hadoop-qbt-trunk-java8-linux-x86_64/290/testReport/junit/org.apache.hadoop.crypto.key.kms.server/TestKMS/testACLs/ Please give a check > LoadBalanceKMSClientProvider#deleteKey should invalidateCache via all > KMSClientProvider instances > - > > Key: HADOOP-17208 > URL: https://issues.apache.org/jira/browse/HADOOP-17208 > Project: Hadoop Common > Issue Type: Improvement >Affects Versions: 2.8.4 >Reporter: Xiaoyu Yao >Assignee: Xiaoyu Yao >Priority: Major > Labels: pull-request-available > Fix For: 3.4.0 > > Time Spent: 2h 20m > Remaining Estimate: 0h > > Without invalidateCache, the deleted key may still exists in the servers' key > cache (CachingKeyProvider in KMSWebApp.java) where the delete key was not > hit. Client may still be able to access encrypted files by specifying to > connect to KMS instances with a cached version of the deleted key before the > cache entry (10 min by default) expired. -- This message was sent by Atlassian Jira (v8.3.4#803005) - To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org