Re: [SSI-users] kernel: reop_export_path

[John B. <joh...@hp...>]

Andy Phillips wrote:
> Hi,
> 
>    I've built an Open SSI cluster, based around 3 dual Xeon 1 u servers.
> I've done a vanilla install of fedora core 2, and the 2.4 based 
> OpenSSI 1.2.2 release.
> 
> The one change I've made was to roll my own kernel with the latest
> qlogic fibre card drivers and intel e1000, based off the i686-smp
> config. 
> 
>    I'm getting a lot of messages like
> May 27 18:09:38 logssi1 kernel: reop_export_path: Can't export unlinked
> file /cluster/dev/pts/4 (deleted)
> 
>    And checking over the /cluster/dev/pts gives,
> total 0
> crw--w----  1 atp tty 136, 0 May 31 14:23 0
> ?---------  ? ?   ?        ?            ? 1
> crw--w----  1 atp tty 136, 2 May 27 11:39 2
> crw--w----  1 atp tty 136, 3 May 31 14:03 3
> ?---------  ? ?   ?        ?            ? 4
> crw--w----  1 atp tty 136, 5 May 31 14:16 5
>    
>    I saw an email on this list a dated this time last year that said
> that this was a known issue, and that there was a fix. Was the fix
> applied and is this a serious problem. I can still log into the system,
> but these pty's are not being re-used.
> 
>    What may be happening is that the sshd is not going down cleanly,
> as an intervening firewall will occasionally terminate idle ssh
> connections. 

Even if it goes down uncleanly, things should still be cleaned up when 
the sshd terminates. Which is should do, eventually. I'll try this, though.

Does "lsof | grep pts" show the problem pts devices anywhere?

> 
>    Any ideas? Is this an old problem, and theres a patch I've missed?
> Or can I safely ignore this?


This is mostly just annoying as far as I know; at least until all the 
available ptys get used up. If I can reproduce the problem, I will 
provide a fix.

John Byrne