Yifan Zhang has posted comments on this change. ( http://gerrit.cloudera.org:8080/14420 )
Change subject: KUDU-1620: Update consensus peer proxies when a network error occurs ...................................................................... Patch Set 3: (2 comments) http://gerrit.cloudera.org:8080/#/c/14420/1/src/kudu/consensus/consensus_peers.cc File src/kudu/consensus/consensus_peers.cc: http://gerrit.cloudera.org:8080/#/c/14420/1/src/kudu/consensus/consensus_peers.cc@518 PS1, Line 518: consensus_proxy_(std::move(DCHECK_NOTNULL(consensus_proxy))) { > RPC calls wouldn't need to be more time consuming per se. Ideally the very KUDU-75 says "do DNS resolution inline with calls", I thought it means resolve the address (or use cache) each time we send a RPC request. If store the Sockaddr and only on errors retry to do DNS resolution, we also need to add special handling in callback functions. http://gerrit.cloudera.org:8080/#/c/14420/1/src/kudu/consensus/consensus_peers.cc@524 PS1, Line 524: const rpc::ResponseCallback& callback) { > I see you added a lock, but I still don't understand the requirements. Coul I'm not sure if multiple threads end up Peer::ProcessResponse, if not, the lock is not required. -- To view, visit http://gerrit.cloudera.org:8080/14420 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-Project: kudu Gerrit-Branch: master Gerrit-MessageType: comment Gerrit-Change-Id: Ied476d302b1d9489a030847b5da74d60d0c1d8e9 Gerrit-Change-Number: 14420 Gerrit-PatchSet: 3 Gerrit-Owner: Yifan Zhang <[email protected]> Gerrit-Reviewer: Adar Dembo <[email protected]> Gerrit-Reviewer: Kudu Jenkins (120) Gerrit-Reviewer: Yifan Zhang <[email protected]> Gerrit-Comment-Date: Tue, 15 Oct 2019 06:48:17 +0000 Gerrit-HasComments: Yes
