I/O Stall Patches

[ByteEnable]

Hi Hannes, Mike.

I've noticed that Hannes has been working a I/O stall issue and has
created some patches. I'm curious because I'm seeing some I/O stall
when I'm logged in to multiple targets. Is there a way to detect the
signature of the I/O stall which Hannes is fixing?

Thanks.

[Mike Christie]

What type of stall are you seeing?

In /var/log/messages do you see something about a iscsi nop/ping timing
out, or do you see something about a target or host or lun reset
succeeding/failing?

[ByteEnable]

I'm seeing ping time out's with an occasional tur failure from
multipath which in turn kills the session on the path that fails. No
TMF stuff.

Byte

[Mike Christie]

What version of open-iscsi and what is the ping timeout?

Could you try the kernel modules and tools from
http://www.open-iscsi.org/bits/open-iscsi-2.0-871.tar.gz. I did a tiny
change to the ping code, and it looks like for some other group it has
fixed their problem (at least I have not heard back from them in a
couple of weeks).

[ByteEnable]

On Aug 26, 9:09 pm, Mike Christie <micha...@cs.wisc.edu> wrote:
> On 08/26/2009 07:58 PM, ByteEnable wrote:
>
>
>
>
>
> > On Aug 26, 4:19 pm, Mike Christie<micha...@cs.wisc.edu>  wrote:
> >> ByteEnable wrote:
> >>> Hi Hannes, Mike.
> >>> I've noticed that Hannes has been working a I/O stall issue and has
> >>> created some patches.  I'm curious because I'm seeing some I/O stall
> >>> when I'm logged in to multiple targets.  Is there a way to detect the
> >>> signature of the I/O stall which Hannes is fixing?
> >> What type of stall are you seeing?
>
> >> In /var/log/messages do you see something about a iscsi nop/ping timing
> >> out, or do you see something about a target or host or lun reset
> >> succeeding/failing?
>
> > I'm seeing ping time out's with an occasional tur failure from
> > multipath which in turn kills the session on the path that fails.  No
> > TMF stuff.
>
> What version of open-iscsi and what is the ping timeout?
>

> Could you try the kernel modules and tools fromhttp://www.open-iscsi.org/bits/open-iscsi-2.0-871.tar.gz.  I did a tiny

> change to the ping code, and it looks like for some other group it has
> fixed their problem (at least I have not heard back from them in a
> couple of weeks).

This is on RHEL5U4 first or so beta I believe.

[ByteEnable]

It's just a regular ping timeout. I'm not in front of the console but
if I remember correctly its the standard timeouts set in iscsid.conf

[Mike Christie]

Ah ok, then try this kernel
http://people.redhat.com/dzickus/el5/164.el5/