-
Type: Bug
-
Resolution: Done
-
Priority: Critical - P2
-
None
-
Affects Version/s: None
-
Component/s: WiredTiger
-
None
-
Environment:Debian Wheezy.
MongoDB 3.0.7
-
Linux
On a daily basis we have to restart MongoDB nodes (secondaries and primary) because performance degrades dramatically.
We can observe the percentage of WiredTiger cache (%use column when we use mongostat command) in use are freezing to 100% (and this percentage never decrease by it's own if we don't restart the node).
After the node has been restarted, the %use decrease to ~80% generally.
I've added an attachment of a graph displaying the %use on our servers.
We tried to increase the cacheSize parameter.
Here on a 64GB server:
wiredTiger:
engineConfig:
cacheSizeGB: 56
The issue is still the same even with this configuration.