Todd Lipcon has uploaded this change for review. ( http://gerrit.cloudera.org:8080/9375
Change subject: WIP: KUDU-2297 (part 5): dump stacks into diagnostics log on service queue overflow ...................................................................... WIP: KUDU-2297 (part 5): dump stacks into diagnostics log on service queue overflow In the past we've had several bugs and performance issues which manifest themselves as service queue overflows. It would be easier to pinpoint the root cause (eg slow IO vs lock contention) if we had a process-wide stack snapshot at the time of the overflow. This patch does some plumbing to trigger such a stack trace into the diagnostics log when a service queue overflow occurs. The logging is throttled to once every 5 seconds so that we don't contribute too much to the load itself. The plumbing is a bit ugly since the diagnostics log is in the server/ module whereas the overflows surface in the rpc/ module. Having rpc/ call back into server/ would be a cyclic dependency, so instead the rpc/ module exposes a std::function<> style callback for the server to hook into. WIP because I'd like to add some kind of test for this, though I tested manually and it appears to work. Change-Id: I62019c71121d7ca4037cab86a7c2ea048a261ad8 --- M src/kudu/rpc/service_pool.cc M src/kudu/rpc/service_pool.h M src/kudu/server/diagnostics_log.cc M src/kudu/server/diagnostics_log.h M src/kudu/server/rpc_server.cc M src/kudu/server/rpc_server.h M src/kudu/server/server_base.cc M src/kudu/server/server_base.h 8 files changed, 96 insertions(+), 11 deletions(-) git pull ssh://gerrit.cloudera.org:29418/kudu refs/changes/75/9375/1 -- To view, visit http://gerrit.cloudera.org:8080/9375 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-Project: kudu Gerrit-Branch: master Gerrit-MessageType: newchange Gerrit-Change-Id: I62019c71121d7ca4037cab86a7c2ea048a261ad8 Gerrit-Change-Number: 9375 Gerrit-PatchSet: 1 Gerrit-Owner: Todd Lipcon <t...@apache.org>
