Details
-
Bug
-
Resolution: Done
-
Major - P3
-
None
-
3.0.3
-
None
-
ALL
-
Description
We have a replica set with one secondary and one arbiter (each on a different server) and moved from MMapv1 to WiredTiger. Primary and Secondary each have 32 GB. We have now a process where we read each document once. The mongod process uses all of the 32 GB RAM and the server is very slow.
Attached you will find our serverStatus: wt cache size is 16 GB, tcmalloc size is 26 GB.
So the question for us is, why WT uses all of the 32 GB and swaps now and how can we fix that. Also, if we upgrade to a new version, we see in the jira tasks, that there are a lot of problems with memory leaks together with WT. So what recommendation do you have.
free -m output:
total used free shared buffers cached
Mem: 32136 31727 409 0 9 2899
-/+ buffers/cache: 28819 3317
Swap: 49152 609 48543