[xcat-user] Failed to allocate pty

[Egan Ford]

Did a reboot fix it?  Can you post the exact text of the error?
 
Thanks.

-----Original Message-----
From: xcat-user-bounces at lists.xcat.org
[mailto:xcat-user-bounces at lists.xcat.org] On Behalf Of Antonis A.
Constantinou
Sent: Thursday, April 10, 2008 9:59 AM
To: xCAT Users Mailing list
Subject: RE: [xcat-user] Failed to allocate pty
Importance: High



I am using RHEL 5 x 64. It really strange. Still the problem is there. 

 

Antonis

 


  _____  


From: xcat-user-bounces at lists.xcat.org
[mailto:xcat-user-bounces at lists.xcat.org] On Behalf Of William Marshall
Sent: Thursday, April 10, 2008 6:33 PM
To: xCAT Users Mailing list
Subject: Re: [xcat-user] Failed to allocate pty

 


xcat-user-bounces at lists.xcat.org wrote on 04/10/2008 09:57:20 AM:

> All of a sudden I can not login in my masternode. IT started 
> yesterday night to give me Failed to allocate pty error when I tried
> ssh to the masternode .Authentication works as I can see that I get 
> passed authentication in the ssh debug log.SSH login on the nodes 
> works fine.I have compared the ssh logfiles and are exactly the 
> same. Just the masternode when tries to open the shell it says 
> failed to allocate pty. 
>   
> I have checked /dev/pts and it is mounted.Also no space issues on 
> partitions.Restarting masternode did not helped. Restarting sshd 
> also did not helped. 

We've seen something similar, but we could reboot the system to get sshd
working again. What distro are you running? 

I had an IBM PMR 09885,379,000 open on this in Feb. I don't know if you can
see that and/or can open an IBM PMR. They closed it w/ at this point the
recommendation would be to upgrade to at least RHEL 4 U5 to get those fixes
and see if the issue persists.  In all likelyhood we would have to reproduce
this on RHEL 4 U6 to take any sort of issues to redhat. 

Here's more details from the PMR: 

After speaking with our L3 team and walking through the strace, I've got
some new information for you. 

Looking over the strace, we see an I/O error while opening the /dev/ptmx
device. 
10454 open("/dev/ptmx", O_RDWR <unfinished ...> 
10454 <... open resumed> )              = -1 EIO (Input/output error) 


Here is the bit of code that returns the above error:

idr_ret = idr_get_new(&allocated_ptys, NULL, &index);
if (idr_ret < 0) {
up(&allocated_ptys_lock);
if (idr_ret == -EAGAIN)
return -ENOMEM;
return -EIO;
}
if (index >= pty_limit) {
idr_remove(&allocated_ptys, index);
up(&allocated_ptys_lock);
return -EIO;
}

  

 


 

 

 


Basically what happens is as we are trying to allocate a new pty, we walk
through an index of used and available ptys.  Somewhere in there we are
blowing up trying to find an available device.  The index record table may
have gotten messed up somehow. 

Looking at the change log from the kernel there are some fixes in a later
kernel version which address issues with freeing indexes. 


* Tue Jul 24 2007 Jason Baron <jbaron at redhat.com> [2.6.9-55.22]
-tty_io: grab tty_sem while changing tty struct in release_dev (Aristeu
Rozanski
) [230747]

* Tue Mar 20 2007 Jason Baron <jbaron at redhat.com> [2.6.9-51]
-tty: fix possible double idr index freeing (Aristeu Rozanski) [214188]




Bill Marshall
Integrated Technology Delivery, Server Operations
Rochester PC Server Team
Building 020-2, Rochester, MN 



-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.xcat.org/pipermail/xcat-user/attachments/20080410/7c77a9e9/attachment.htm