Details
-
Bug
-
Resolution: Fixed
-
Critical
-
2.5.1
-
Security Level: Public
-
The database is composed of three db server nodes: 172.23.96.64 172.23.96.65 172.23.96.62
Replication is turned off during the testing.
-
Untriaged
-
Centos 64-bit
-
Unknown
Description
The output from sync-gateway's expvar (http://172.23.96.63:4985/_expvar) shows that a large number of tap responses from Couchbase db takes over 1 minute. For example:
"lag-total-28900ms": 8676, (8676 tap responses took 28 seconds)
"lag-total-56400ms": 21,
"lag-total-20300ms": 17632,
"lag-total-26900ms": 6564,
"lag-queue-49100ms": 45,
This happen when two sync-gateways, each handles 10K users, are running in parallel. This happen about one hour after the test started.
Attached are the two couchbase console screen shots, and the expvar from sync_gateway that shows the above tap responses.
The cbcollect_info zip file is too large to attach to this ticket. It can be accessed from http://172.23.106.228/20140424-164336-8cpu8g-3gw3gl3db-u10000-c16-fail/cblog-64.zip
Attachments
Issue Links
- relates to
-
MB-11037 High replication latency due to the racing in pausing / resuming TAP connection notifier
- Closed