Details
-
Bug
-
Resolution: Fixed
-
Critical
-
5.5.0
-
5.5.0-2017; CentOS longevity (see
MB-28453).
-
Untriaged
-
Unknown
Description
When investigating MB-28453, I found that the memcached logs wrapped early (after 4H) due to the volume of log messages; and MB-28456 was opened to investigate.
Amongst legitimate excessive logging by the memcached process (which we are fixing now); there were also "legitimate" warnings where FTS was attempting to Stream non-existenct vBuckets:
$ grep "WARNING (default) DCP (Producer) eq_dcpq:fts" memcached.log |grep "Stream request failed because this vbucket doesn't exist" | cut -f 2- -d ' ' |wc -l
|
61282
|
i.e. 61,282 instances over 4 hours; average of ~255 per minute.
This seems to be pointing at a problem in FTS - I wouldn't expect it would need to poll for Streams this way (should be driven by the vBucket map).
Could you take a look and see if this is a problem on the FTS side?