Tweet
Within a cP user account I can see the stats, which are averaged over a minute (as granular as the cP report gets), which are minimal (4-6 processes) but through ssh Im seeing spikes in activity every five minutes for just a couple of seconds during which processes are unable to fork, I assume due to LVE limits (either EP or NPROC) being exceeded. There are no "faults" reported in cP user console.
Ive confirmed that there are no cron jobs in the user crontab, there is no activity to the website from the Internet, and no dovecot sessions from mail clients etc. I must assume that there are a few processes I cant see through ps -ef related to the user account, such a mail processing, httpd, etc but these should equally be pretty idle.
Via ps -ef I only every see around half a dozen processes max. Via /proc/stat the number of processes can vary from the same half dozen up to 50+, but an ls of /proc only shows half a dozen process directories. I can only assume that ps only shows procs spawned directly under the uid, and not those that have set an effective uid. Similarly /proc has similar view. I assume /proc/stat shows the real process count of all processes with the uid.
I can accept that there will be some activity, httpd, backups etc, but these background activities are almost hitting limits; either EP or NPROC, or both, but Im struggling to diagnose which. If I run just a few more processes from the shell, it reports an unable for fork message on the 5 minute boundaries.
So, are my assumptions about ps and /proc/stat correct? Is there a way from within a standard end user account / cagefs to obtain a true view of EP and NPROC. Is there a way see the additional processes (not shown by ps)?
Im basically trying to figure out what the true counts are, and why the limits are being (almost) hit before any real use of the account is taking place. Any insights would be much appreciated.
Ive confirmed that there are no cron jobs in the user crontab, there is no activity to the website from the Internet, and no dovecot sessions from mail clients etc. I must assume that there are a few processes I cant see through ps -ef related to the user account, such a mail processing, httpd, etc but these should equally be pretty idle.
Via ps -ef I only every see around half a dozen processes max. Via /proc/stat the number of processes can vary from the same half dozen up to 50+, but an ls of /proc only shows half a dozen process directories. I can only assume that ps only shows procs spawned directly under the uid, and not those that have set an effective uid. Similarly /proc has similar view. I assume /proc/stat shows the real process count of all processes with the uid.
I can accept that there will be some activity, httpd, backups etc, but these background activities are almost hitting limits; either EP or NPROC, or both, but Im struggling to diagnose which. If I run just a few more processes from the shell, it reports an unable for fork message on the 5 minute boundaries.
So, are my assumptions about ps and /proc/stat correct? Is there a way from within a standard end user account / cagefs to obtain a true view of EP and NPROC. Is there a way see the additional processes (not shown by ps)?
Im basically trying to figure out what the true counts are, and why the limits are being (almost) hit before any real use of the account is taking place. Any insights would be much appreciated.
Comment