Terminal id's not being reset 1

[davesa]

We have a server running Fedora, ver 2.6. The terminal ID's aren't being released, so that when pts/1, pts/2 and pts/3 are logged in and pts/1 logs out, the next person to log in gets pts/4 instead of pts/1.

Is there a setting somewhere we have missed?

Please help...

 

[ericbrunson]

Does ps or lsof show anything using those ports?

 

[davesa]

I ran lsof and ps and the only reference to the running app is for the terminals that show up in the who list.

 

[marsd]

Those are not true terminal but pseudo terminal devices.
Are you getting any error messages via syslog?
If you close all pty sessions and login again do you
get pty1?
More info needed.

 

[stoffelsa]

there is no error msg. in the syslog file (etc/sysconfig).
The actual problem occurs when a dumb term. logs in via the network,so then it would get an pseudo no. (/dev/pts/0-?), but in this case the pseudo devices just goes on and on. It doesn't re-use the logged out ones.

end

 

[ericbrunson]

Do you ever run out of pseudo terminals?

And when I asked about lsof, did you try "lsof /dev/pts/x" where x = a pts that didn't get freed? When the pts is freed the device node should go away, if not then lsof should show you what's on it.

 

[thedaver]

I'm curious, is the anticipated "default" behavior to always try the "first" numerically available pts or is it to use the "next scheduled" a la a round-robin DNS thing? I'm just wondering if this is a red herring question. I don't have the foggiest idea, but linux doesn't strike me as having a bias to always returning to the lowest numbered available anything when drawing from a pool of like resources.... I may be all wet too.....

Hosting Solutions for Home or Business.

http://www.surfinbox.com/hosting/

 

[stoffelsa]

it doesn't look like it would run out,i saw /dev/pts/30 once.When i do lsof it's only the active/current logged in users pts no.here is a eg. lsof /dev/pts/2 (lsof: status error on /dev/pts/2: No such file or directory)
I am actually thinking of re-installing with FC1 - i know that works.

 

[stoffelsa]

It seems that it is a bug for kernel prior to 2.6.6 but i am not sure that it has been fixed. See this thread (and especially the last message by Andrew Morton) :

http://lkml.org/lkml/2004/4/5/248