[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-02-02 Thread Anoop Sam John (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13569584#comment-13569584 ] Anoop Sam John commented on HBASE-3787: --- +1 on above suggestion.

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-02-01 Thread Ted Yu (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13568926#comment-13568926 ] Ted Yu commented on HBASE-3787: --- Maybe we can create subtasks for this JIRA, beginning with

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-02-01 Thread Andrew Purtell (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13568970#comment-13568970 ] Andrew Purtell commented on HBASE-3787: --- bq. Maybe we can create subtasks for this

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-02-01 Thread Ted Yu (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13569160#comment-13569160 ] Ted Yu commented on HBASE-3787: --- I am currently looking into test failures first reported by

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-02-01 Thread Anoop Sam John (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13569441#comment-13569441 ] Anoop Sam John commented on HBASE-3787: --- [~apurtell] I will try with this on

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-02-01 Thread Lars Hofhansl (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13569451#comment-13569451 ] Lars Hofhansl commented on HBASE-3787: -- The discussions here all imply an RPC and WAL

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-01-31 Thread Ted Yu (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13567807#comment-13567807 ] Ted Yu commented on HBASE-3787: --- bq. but server failed before returning response. Client

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-01-31 Thread Andrew Purtell (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13567864#comment-13567864 ] Andrew Purtell commented on HBASE-3787: --- I think the above comments all taken

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-01-31 Thread Ted Yu (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13567872#comment-13567872 ] Ted Yu commented on HBASE-3787: --- bq. We should append the nonce to the WALEdit, and recover

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-01-31 Thread Andrew Purtell (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13567874#comment-13567874 ] Andrew Purtell commented on HBASE-3787: --- bq. Is the above needed ? I think Enis is

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-01-31 Thread Ted Yu (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13567877#comment-13567877 ] Ted Yu commented on HBASE-3787: --- bq. without having a nonce the increment would be accepted

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-01-31 Thread Andrew Purtell (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13567890#comment-13567890 ] Andrew Purtell commented on HBASE-3787: --- Good point Ted. So then the client should

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-01-31 Thread Andrew Purtell (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13567897#comment-13567897 ] Andrew Purtell commented on HBASE-3787: --- What we really need as a different model

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-01-31 Thread Ted Yu (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13567896#comment-13567896 ] Ted Yu commented on HBASE-3787: --- If client retries on region move, that would allow skipping

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-01-31 Thread Andrew Purtell (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13567930#comment-13567930 ] Andrew Purtell commented on HBASE-3787: --- [~ted_yu] From the client's point of view,

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-01-31 Thread Ted Yu (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13567944#comment-13567944 ] Ted Yu commented on HBASE-3787: --- For statement #1: bq. HBaseClient should generate a new

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-01-31 Thread Andrew Purtell (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13568022#comment-13568022 ] Andrew Purtell commented on HBASE-3787: --- How does the client know if the op failed

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-01-31 Thread Ted Yu (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13568043#comment-13568043 ] Ted Yu commented on HBASE-3787: --- clarification: bq. HBaseClient should generate a new nonce

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-01-30 Thread Andrew Purtell (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13566821#comment-13566821 ] Andrew Purtell commented on HBASE-3787: --- I think this has come up again on the user

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-01-30 Thread Ted Yu (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13566840#comment-13566840 ] Ted Yu commented on HBASE-3787: --- So the server would apply Increment requests with the same

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-01-30 Thread Andrew Purtell (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13566847#comment-13566847 ] Andrew Purtell commented on HBASE-3787: --- bq. So the server would apply Increment

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-01-30 Thread Enis Soztutar (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13566915#comment-13566915 ] Enis Soztutar commented on HBASE-3787: -- This is a bit orthogonal, but we need a

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-01-30 Thread Andrew Purtell (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13566936#comment-13566936 ] Andrew Purtell commented on HBASE-3787: --- If we define the nonce as per-op instead of

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-01-30 Thread Ted Yu (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13567136#comment-13567136 ] Ted Yu commented on HBASE-3787: --- On region server, we need to keep a map from row key to set

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-01-30 Thread Andrew Purtell (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13567168#comment-13567168 ] Andrew Purtell commented on HBASE-3787: --- bq. On region server, we need to keep a map

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-01-30 Thread Ted Yu (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13567177#comment-13567177 ] Ted Yu commented on HBASE-3787: --- bq. (client-address, nonce, timestamp) should be

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-01-30 Thread stack (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13567180#comment-13567180 ] stack commented on HBASE-3787: -- I like this idea. Increment is

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-01-30 Thread Andrew Purtell (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13567183#comment-13567183 ] Andrew Purtell commented on HBASE-3787: --- bq. We haven't discussed how the nonce

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-01-30 Thread Ted Yu (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13567257#comment-13567257 ] Ted Yu commented on HBASE-3787: --- bq. I was thinking hash(IP address, row, timestamp). Should

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-01-30 Thread Anoop Sam John (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13567322#comment-13567322 ] Anoop Sam John commented on HBASE-3787: --- HBASE-6390 is the same issue.

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-01-30 Thread Andrew Purtell (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13567338#comment-13567338 ] Andrew Purtell commented on HBASE-3787: --- Closed HBASE-6390, older one wins.

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-01-30 Thread Anoop Sam John (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13567382#comment-13567382 ] Anoop Sam John commented on HBASE-3787: --- [~apurtell] I am trying to understand the

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-01-30 Thread Enis Soztutar (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13567401#comment-13567401 ] Enis Soztutar commented on HBASE-3787: -- BTW, having an in-memory windowing view of

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2013-01-30 Thread Anoop Sam John (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13567410#comment-13567410 ] Anoop Sam John commented on HBASE-3787: --- [~enis] Yes you bring out an extreme case.

[jira] [Commented] (HBASE-3787) Increment is non-idempotent but client retries RPC

2012-09-13 Thread Lars Hofhansl (JIRA)
[ https://issues.apache.org/jira/browse/HBASE-3787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13455384#comment-13455384 ] Lars Hofhansl commented on HBASE-3787: -- This is an interesting one. To me the main

<    1   2