Restart of mongod pods - WiredTiger Cache, Percona Memory Engine

I was wondering what happens if a mongod pod gets restarted (for ex. simple liveness probe failure or moved to another node … both viable reasons) and 100% of the memory is used for WT Cache and/or Percona Memory Engine.

Is it ensured that no data is lost ?

Further when the pod starts up then it takes very long until these checkpoints are reached. We had to set like

livenessProbe:
  initialDelaySeconds: 300
  failureThreshold: 10

until mongod was able to start. Now the disk is like 5% full - I just wonder what happens if the disk is full - most likely it will take like 1h.

Is there a possibility to make the (re)start much faster ?

1 Like

It looks like mongod always gets shutdown the “hard” way …

{“t”:{"$date":“2021-07-14T23:15:06.750+00:00”},“s”:“W”, “c”:“STORAGE”, “id”:22271, “ctx”:“initandlisten”,“msg”:“Detected unclean shutdown - Lock file is not empty”,“attr”:{“lockFile”:"/data/db/mongod.lock"}}

1 Like

https://jira.mongodb.org/browse/SERVER-43664

1 Like