Open rueberger opened 6 years ago
Hey @rueberger !
I'm working around this issue too, but I still looking for a workaround, without success. Perhaps, do you had found the way to fix this behaviour?
My lab on Kubernetes: Sharding Cluster: 1 replica set Config Servers (3 servers), 1 Shard (1 replica set, 3 servers + 2 arbiters) and 2 mongos.
Thanks, Ronald E.
I haven't had any issues for a while, but I'm not really sure what changed. Now running mongo:3.6
, with the following mongo.conf
:
storage:
dbPath: /data/db
wiredTiger:
engineConfig:
cacheSizeGB: 6
and a container memory limit of 10G. Hope that helps.
Apologies if this isn't the best place for this issue.
I've noticed that with docker 17.12.0-ce and the
mongo:3.4.1
image, memory limits imposed on the container are not "respected" by mongo. What I mean by this is that mongo seems to not recognize the memory limit and tries to use the full memory of the host system.This leads to bad things happening when memory fills up - mongo starts spilling into swap and performance slows to a crawl.
I would rather not give mongo unrestricted access to the host's memory, or be forced to periodically kill mongod.
I considered explicitly limiting the cache size for WiredTiger, but that's no good because the file-system cache is still unbounded.
Any ideas for a workaround?