Dear all.
I have a machine running on RH with 4G memory and 8G swap.
Just now I'm having a problem when loging in with a user - just call it "userX" -.

Each time I'm logged in I got a message
And almost everytime I entered any command, it gives me the same warning. In some time (rarely) it will give me proper result, but almost any time it gives me the warning and no result.

When I use 'top' command, it shows that I still have 500M free memory and about 8G free swaps. Plus the process for userX is less than 20 processes, and each of them are short processes (less than an hour, mostly just few seconds).

As a note this problem only occur on userX, for any other user on the machine everything works just fine.

PS. my box is a web server, and mostly CPU load about 50%.

Any ideas welcome


Regards,

YoChan

Check the output of ulimit -a
This will list the number of processes/files you have start/have open. You are likely running into this number.

Also see /etc/security/limits.conf and /etc/security/limits.d/ for resource limits configured by pam_limits.so per user. In this file you can configure the particular user's limits higher, but you have to logout and in for this to take effect.

1 members found this post helpful.

There are other resources besides memory. The number of open files. The number of locks. The maximum filesize. The maximum stacksize.

You may want to check the limits.conf file as already suggested. Also check the files in the users home directory. I've had ~/.xsession errors grow to an extremely large size due to one of earlier KDE 4 programs.

Also read your kernel log. It probably will indicate exactly which resource limit was reached.

1 members found this post helpful.

Thanks for the replies.
I think it answered my question.

Only 20 processes per user.... (sigh)

Currently the issue stopped, but still should I change it to a bigger number?

Do you regularly use more than 20 processes per user? Yes? Yes.

1 members found this post helpful.

That greatly depends on who has accounts on the box, and how it's used.

If you're the sysadmin, and you don't run any automated jobs that may run amuck, I'd sys up it to something reasonable. For a box that has a lot of concurrent console users 20 actually isn't unreasonable, but it could be low for "real" work. If it's console only, maybe up it to 64, if you use a graphical desktop on it maybe 128. If you're writing or using code that forks a lot, you could go 64k. Heck, you could go unlimited, but if this is a production box, that means anything run with your account could kill the system with a forkbomb, etc.

Do you know why it's set so low currently?

1 members found this post helpful.

Sorry for my late reply, and thanks again for the replies.

@AlucardZero: Sometimes, YES. Weekly one of my user use webalizer, and opens it several instance. Each instance eats one process. When she do her job, there will be no spare process left for me to do some work for the server maintenance. But as long as I stay away from the server than it's all fine

@kschmitt: I don't know when, and how. It's an artifact server. I think (IMHO) 20 is way too small. But in most cases we don't work directly in this server for it's a web server. And web request (AFAIK) handled by user "nobody" so basically as I said to AlucardZero 20 is enough as long as no one lay his feet on the server while it's working it's regular job.

PS. can anyone tells where can I read my kernel log? I want to check what it said about this problem.