Gabriella Lotz has uploaded this change for review. ( 
http://gerrit.cloudera.org:8080/24594


Change subject: auto_leader_rebalancer: best-effort leader step-down
......................................................................

auto_leader_rebalancer: best-effort leader step-down

When rebalancing leaders for a table, RunLeaderRebalanceForTable
resolved the current leader's address and sent a LeaderStepDown RPC
with RETURN_NOT_OK. A single transient failure (an RPC timeout, the
target briefly unavailable, or an election in flight) aborted the whole
rebalancing pass and skipped every remaining tablet and table in that
round.

A leader transfer is best effort. The next round recomputes the plan
and retries, so a single failed transfer should warn and move on to the
next tablet rather than fail the pass. RunLoop already only logs a
warning for the overall call, so this just stops one bad RPC from
starving the rest of the round.

Seen under heavy load, where one failed LeaderStepDown would otherwise
end a round early and leave the leaders unbalanced for longer.

Change-Id: I0fb2c7a8bf750520d34d9aba62274dbe7433d80d
---
M src/kudu/master/auto_leader_rebalancer.cc
1 file changed, 14 insertions(+), 2 deletions(-)



  git pull ssh://gerrit.cloudera.org:29418/kudu refs/changes/94/24594/1
--
To view, visit http://gerrit.cloudera.org:8080/24594
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-MessageType: newchange
Gerrit-Change-Id: I0fb2c7a8bf750520d34d9aba62274dbe7433d80d
Gerrit-Change-Number: 24594
Gerrit-PatchSet: 1
Gerrit-Owner: Gabriella Lotz <[email protected]>

Reply via email to