Re: Large number of ssh_poll/poll leads to "Interrupted system call"

[Andreas Schneider <asn@xxxxxxxxxxxxxx>]

On Wednesday 10 September 2014 15:28:31 Dennis, Brent - 1008 - MITLL wrote:
> Hi,

Hi Brent,
 
> I have an application which is trying to execute ~100-200 remote commands on
> multiple servers via libssh.  I can control the rate at which these
> commands are invoked (currently every ~5 ms).  Occasionally, the poll
> invoked by the ssh_poll() (rooted itself with an ssh_channel_open_session()
> ) returns with the "Interrupted system call" error, which ultimately
> corrupts my ssh_session such that subsequent ssh_channel_open_session()
> will also error out.
> 
> I tried to catch the signal that was interrupting the poll() but have been
> unsuccessful.  The problem is intermittent, but once it occurs, the session
> must ultimately be rebuilt.
> 
> Does anyone have any insight as to what could be happening?  Any ideas on
> how to narrow down determining the event which interrupts for the poll()? 
> Or are there any mitigation steps to take should this type of event occur?

Sorry for the late reply. We have no idea. However it would be nice to have 
runtime detection for epoll() and use it if available on Linux. I need to 
check how we implemented it in tevent.


Cheers,

	-- andreas

-- 
Andreas Schneider                   GPG-ID: CC014E3D
www.cryptomilk.org                asn@xxxxxxxxxxxxxx