Todd Lipcon has posted comments on this change. Change subject: KUDU-1806. java: fetching scan tokens should fetch larger batches ......................................................................
Patch Set 3: OK, here are results from YCSB: - ran a 'load workloada' with sync_ops = true to cause a random write workload, while I captured a trace with /tracing.html. I kill -STOPped a server to cause some leader elections, which caused the client to re-fetch locations. I verified that all of the GetTableLocations RPCs in the trace had max_returned_locations=10 - I did the same thing for 'run workloadc' which is a read-only workload (range scans) - I wrote a little script to call Kudu.getTabletsLocations(null, null, <timeout>) directly, and verified that the trace showed a single call with max_returned_locations=1000 This was all with a version of YCSB that uses range partitioning. I'll try upgrading to the latest version from git which uses hash partitioning for one more check before I push. -- To view, visit http://gerrit.cloudera.org:8080/5498 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-MessageType: comment Gerrit-Change-Id: I46260a96dfd0847f70146496e48c2766b8e17ea9 Gerrit-PatchSet: 3 Gerrit-Project: kudu Gerrit-Branch: master Gerrit-Owner: Todd Lipcon <[email protected]> Gerrit-Reviewer: Adar Dembo <[email protected]> Gerrit-Reviewer: Dan Burkert <[email protected]> Gerrit-Reviewer: Jean-Daniel Cryans <[email protected]> Gerrit-Reviewer: Kudu Jenkins Gerrit-Reviewer: Matthew Jacobs <[email protected]> Gerrit-Reviewer: Todd Lipcon <[email protected]> Gerrit-HasComments: No
