This can cause significant problems for Couchbase Server as it typically needs more threads than that (say on a 16 core machine) - see linked CBSEs.
As I understand it, the Operator cannot directly check the limit; however I believe that the limit is exposed inside the container via /sys - specifically /sys/fs/cgroup/pids/pids.max:
Request is to check the value of this, and fail certification if the value is less than some more sensible number - 10,000 seems reasonable given that's what we currently check for the nproc ulimit:
NOTE: I don't have an OpenShift environment to hand, so I cannot be sure if /sys/fs/cgroup/pids/pids.max is the correct path (or even exists there) - that does exist under Docker, and is a reasonable sibling path to the /sys/fs/cgroup/pids/pids.current path mentioned in the aforementioned RH Knowledgebase article.
|Fix Version/s||not-targeted [ 16613 ]|