> Francois,
> If you've solved this problem, please disregard this message.  I've
> marked
> this issue as off topic because it relates to Linux and not Tomcat.

Well, counting FDs is linux related, but why is there so much FDs opened could be 
tomcat related (even if, as a bofh, I would say it's something to do with those thing, 
you know? That specific sort of users called developers). Anyway.
 
> lsof won't properly report the total number of opened file
> descriptors.  I was
> doing the same thing for a client who ran into this issue with file-
> max set to
> 16384.  I was using lsof and wc to count the lines, however it
> reported
> something like 80,000.

        Yes, I ran into this lately yesterday (GMT), lsof reported a number of FDs of 
more than 220000, so as already said, it should report shared FDs as well. Please 
apologize.
 
> You sound pretty knowledgeable, so you may have seen this, however the
> best
> way to handle this is using the proc file system.  An explaination is
> available at
> http://www.ussg.iu.edu/hypermail/linux/kernel/0208.2/0145.html

Yep, I use it.

> Can you post the contents of the following files:
> /proc/sys/fs/file-max
As already mentioned :209713
> /proc/sys/fs/file-nr
2332    1398    209713
Sooooo. 2332 FDs opened, 1398 used. I found a message talking about a hard coded limit 
of 1024 in the Jvm... Is it a per thread limit or ?

> /proc/sys/fs/inode-nr
164972  39841
It seems to be OK there, about 40000 inodes free
> /proc/sys/fs/inode-state
165000  39823   0       0       0       0       0
Show about the same, just adding "everything is OK, good luck, and see you later" in a 
OS point of view.

> Once you've pasted this information, we'll see if we can figure this
> out.

So. I think I ran into a Jvm limit (user is unlimited, shame on me). I will go back to 
the developers corner to ask.

Thank you for your help.

Fran�ois.


---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to