Sailesh Mukil has posted comments on this change. Change subject: IMPALA-5537: Retry RpcRecv on recv timeout exception for SSL connection ......................................................................
Patch Set 1: (1 comment) http://gerrit.cloudera.org:8080/#/c/7229/1//COMMIT_MSG Commit Message: PS1, Line 24: After this change, we will still fail the RPC if there is : any error in the send part of a RPC with secure connection. I'm still slightly worried about this. Our client cache could previously potentially have connections that were already broken for a variety of reasons. Looking at the code now, it seems like it shouldn't be possible, because if an RPC fails, the client_is_unrecoverable_ is set to 'true' and it gets destroyed, but I'm not sure if there is some edge case where a broken connection would be put back into the cache, or worse, a connection gets broken while in the cache (due to a timeout or something similar). If we get a broken connection from the cache, we would fail the RPC without even trying it, as our code would view that as a send() failure. I'll look at it in some more detail to see if and how the above case could happen. -- To view, visit http://gerrit.cloudera.org:8080/7229 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-MessageType: comment Gerrit-Change-Id: I8243d4cac93c453e9396b0e24f41e147c8637b8c Gerrit-PatchSet: 1 Gerrit-Project: Impala-ASF Gerrit-Branch: master Gerrit-Owner: Michael Ho <[email protected]> Gerrit-Reviewer: Henry Robinson <[email protected]> Gerrit-Reviewer: Michael Ho <[email protected]> Gerrit-Reviewer: Sailesh Mukil <[email protected]> Gerrit-HasComments: Yes
