> Most probably, pid_max limit has been reached (can be increased via sysctl)
I somewhat doubt it. A typical OSD spawns ~ 100 threads, thus the default value of
/proc/sys/kernel/pid_max (being 32768) is enough for running ~ 300 OSDs (provided the node has
enough RAM).
So the cause might be a bit simpler: the node needs more RAM to run that many OSDs.
It's hard to find out the exact reason without even basic information about the cluster and nodes.
> Most probably, pid_max limit has been reached (can be increased via sysctl)
I somewhat doubt it. A typical OSD spawns ~ 100 threads, thus the default value of kernel/ pid_max (being 32768) is enough for running ~ 300 OSDs (provided the node has
/proc/sys/
enough RAM).
So the cause might be a bit simpler: the node needs more RAM to run that many OSDs.
It's hard to find out the exact reason without even basic information about the cluster and nodes.