Details
-
Bug
-
Resolution: Fixed
-
Critical
-
4.5.0
-
Untriaged
-
Unknown
Description
For memdb stale=false scan latency test, the call to dcp_get_seqnos takes 150ms.
"bucket-1:myindex:timings/dcp_getseqs":"1000005 150260641165920
Average -> 150ms
This is the test run
http://perf.jenkins.couchbase.com/job/secondary-bharath/581/console
On average, it is about 400 request/s. Background mutation for bucket is about 80K/s (spread over 7 KV nodes). Each KV node has 24 cores. memcached cpu utilization is about 250%-500%. Note that there are 8 KV nodes in the cluster, but one KV node has 0 cpu utilization.
Note that for Q2 stale=false test, dcp_get_seqnos is about 2ms, with low mutation rate (3.6K/s)
by_email:timings/dcp_getseqs":"1377028 2318931713071 -
average -> 1.68ms