Details
-
Improvement
-
Resolution: Unresolved
-
Major
-
None
-
0
Description
Context
For the Capella Health Report iniative (https://couchbasecloud.atlassian.net/browse/AV-66825), we would like to track the SDKs that are connected to a cluster. This will allow us to compare against known EoL/EoM SDKs and inform customers that they are using unsupported software.
There are currently 2 ways to get this information however neither are suitable for our use case.
1) Parse the memcached.log file and look for HELO logs that contain the user agent
This is too resource intensive to run on every Capella node all the time, especially for such a small bit of information. Also, relying on logs that can change is brittle.
2) Use the connections stat e.g.
/opt/couchbase/bin/mcstat -u "xxx" -P "xxx" -b travel-sample connections -j |
The problem with this is that internally kv_engine locks all connections so no other operations can proceed. This would have an impact on tail latencies https://couchbase.slack.com/archives/CFJDXSGUA/p1702312432468979
Ask
We would like this information to be exposed in a way that is easy for dp-agent (the agent running on all Capella nodes) to fetch without negatively impacting the performance of Couchbase Server. We don't have a preference for the exact implementation (e.g. prometheus metric labels/refactor existing connections stat).
Attachments
Issue Links
- blocks
-
AV-66825 Loading...