We are encountering an issue with mongo-4.4.1 (4.4.0 also exhibited the issue) running under kubernetes using: https://hub.docker.com/layers/mongo/library/mongo/4.4/images/sha256-31f6433f7cfcd2180483e40728cbf97142df1e85de36d80d75c93e5e7fe10405?context=explore
The problem is that occasionally the server will spin the CPU, and pile up connections running queries that never finish. It appears that most of them are waiting for a lock that is never unlocked, and they are stuck in that state – the actual TCP connections to the server get shut down by the clients, but server does not decrement the connection count. We can't reproduce this in a reliable fashion – it happens every few days.
The previous version of mongo that we were using was 4.0.5 and we no problems there.
If I can collect any other information please let me know (I have a core dump that I generated the backtraces from, but I can't attach that in case it contains user data).
Thanks,
Mark
- is related to
-
WT-7056 Fix stuck cache eviction worker
- Closed
-
SERVER-54534 No need to wait for writeConcern on aborting a multi-document transaction
- Closed
-
WT-7926 Investigate potential WT cache eviction stuck
- Closed
- related to
-
SERVER-50365 Stuck with long-running transactions that can't be timed out
- Closed
- mentioned in
-
Page Loading...