build 164: latencies vary from 1400 us to 2400 us. Need run newer build > 168 several times to confirm.
not seen in the new runs.
by new runs you mean builds > 168 or something else?
170. looks come back again.
Can you share a couple of reports? Slow run vs. fast run?
right-click and use 'graphs'
Just to double check. This separate ticket is just memcached-side variation right? Nothing to do with rebalance times variations correct ?
Nope, it doesn't have rebalance.
Cluster is being used for toybuild testing will get to this after that.
Lot of variance seen in the rebalance runs, and the mixed large scale tests: <Build 2.0.1-170, 2.0.2-755>
Abhinav, Are you still seeing variances in your latest run with the toy build from Jin?
Can you update this ticket? Thanks.
the issue doesn't show up in 2.0.2 build. will continue to run this for 2.0.2 build
ronnie, can you post the latest perf result from your 2.0.2 run?
how bad the variance with perf on set/get and rebalance in/out?
pls update this bug.
the cluster was constantly engaged with rebalance and mrw medium scale tests. No large scale updates for now. Will schedule them afterwards.
per bug triage, ronnie — can you please update this bug with your large scale test result?
Please refer to mixed-large-4
To verify the variances, need multiple runs of the same 2.0.2 build.
In recent runs the variance delta has reduced. closing this we will reopen if we see high variance.