Hi all,
hoping someone can direct me to where to next look to figure out what I think is a strange issue:

I have a CentOS5 server with the following /etc/security/limits.conf entries that relate to "nproc":
if I run command
it returns 1612 processes (fluctuates but not by more than 10-20).
the server has been online for 220 days and no issues.

Today, if I try and `su - oracle` i get error When I comment out the "soft" limit for nproc (oracle), I can now su without issues.

The fix doesn't makes sense to me as to why it fixed the issue. My questions are:
1) if my ps listing only shows ~1600 PID's for oracle, why would this soft limit be affecting me.
2) total PID's running (all users) is 2044 - coincidence that it is so close to the soft limit?
3) I thought the soft limits were per-process limits (per man), not system-wide limits for a user. In other words, each oracle process is allowed to launch the specified soft limit (up to hard if they change it). Am I mistaken on that?

thanks for any help.

1.
this command should be more honest:
3.
you are correct; it is like this:
new session is opening by UID*→ pam_limits reads /etc/security/limits.conf*→ and sets them with setrlimit()*→ shell clone()s /* EAGAIN may occur here */ and child have the same limits → child execve()s and limits are preserved

child may change EUID to be excluded from parent's limited resources

1 members found this post helpful.

thaks Ygrex - that `ps` command really helped me out.