we're running memcached in containers, with the maximum memory of the process restricted at container startup.
In some cases where the number of keys in the process is big (>100K), our containers die due to Out Of Memory OOM.
We set the memcached memory -m to the expected dataset size, and memory for the container we give about 500MB of slack over the memcached -m. We do this for all containers, but some with high keys number are killed by OOM.
Is the argument for memory (-m) for both keys and values? or just values?
Any idea why these containers go OOM?
Thank you very much.