Monitoring container leaks memory

[IRus]

COS Image: cos-stable-79-12607-80-0
Machine Type: n1-highcpu-2

I running standard GCE VM with COS image, I found out that fluentd eats a lot of RAM without any load on logging (application running in sbx environment, and have just a small amount of startup logs at moment that this behavior discovered)

I logged to instance and check ps -auxf:

ps.txt

In file two outputs from ps, about 20h between runs - memory grows.

[IRus]

Not sure how this possible, docker stats at the same time shows very different numbers:

CONTAINER ID        NAME                        CPU %               MEM USAGE / LIMIT     MEM %               NET I/O             BLOCK I/O           PIDS
14a453ff0f5b        stackdriver-logging-agent   0.33%               150.9MiB / 1.751GiB   8.41%               482kB / 1.4MB       17MB / 3.16MB       28

[sconstable]

I have just seen an issue with out of control memory usage by fluentd on an ubuntu 18.04 VM. Caused our postgres server to run out of memory. fluentd was using 50% of RAM and 100% CPU according to htop.