Suggestions on how to fix fluentd-gcp getting OOMKilled

[Tim Warberg]

Hi

Tried to edit the deamonset but it is almost instantly overwritten by master

Any other options than replacing the auto managed logging with a self-managed? 

--

Tim

[tho...@deepomatic.com]

Hi,
In my kube-up.sh deployed cluster (kubernetes) 1.6 on GCP it is managed by kube-addon-manager which runs on the kubernetes master node, and takes its definitions from /etc/kubernetes/addons/fluentd-gcp.
kube-addon-manager checks changes every minute, and will update the daemonset if you edit the files.

I also have OOMKilled fluend-gcp, even CrashLoopBackOff: the fluentd process eats up its allowed 300MB of memory, and gets killed by the oomkiller. I tried to upgrade to gcr.io/google-containers/fluentd-gcp:2.0.11 but it doesn't change anything...