Am Wednesday, 20. August 2008 19:27 schrieb Mike Christie:
> Klemens Kittan wrote:
> > Here is the whole log.
>
> Thanks. In that log, are you doing io to both disks on both connections
> or just one disks?
>
> It looks like the initiator is not getting a response for the ping in
> time (within ping (node.conn[0].timeo.noop_out_timeout) timeout
> seconds)). It might be a driver bug and we are dropping it. To check
> that you would need to run ethereal or wireshark so we can see what the
> network layer is seeing, but I do not think this is the case given some
> of the other ping times in there.
>
> If the target is just so slow that it is useless to try sending the nop
> you could just run without nops on by setting
> node.conn[0].timeo.noop_out_timeout = 0
> node.conn[0].timeo.noop_out_interval = 0
> or turn node.conn[0].timeo.noop_out_timeout to a higher value like 45
> and see if that helps.
>
> In the logs the responses to some other pings are around 9 seconds
> which is cutting it close when you have a timeout of 10 seconds, so you
> should definately try to increase the node.conn[0].timeo.noop_out_timeout.
>

Hi Mike,

we think now that the problem isn't with the iscsi-stuff but comes
from the ethernet driver (e1000 "Detected Tx Unit Hang"). Sorry and
thank you very much for your patience and help!

Best regards,
Klemens

Attachment: pgpGbUUr6O4rO.pgp
Description: PGP signature

Reply via email to