[OpenIndiana-discuss] Build 147 ZFS / NFS / iSCSI system reboot

[Alasdair Lumsden]

On 18 Feb 2011, at 22:54, Steve Jacobson wrote:

> We're using ietd 1.4.20.2 for the iSCSI target software with a very basic configuration:
> 
> Target iqn.2007-07.com.doyenz:dshelf4.megaraid0
>        Lun 0 Path=/dev/sdc,Type=blockio
>        Lun 1 Path=/dev/sdd,Type=blockio
>        Lun 2 Path=/dev/sde,Type=blockio
>        Lun 3 Path=/dev/sdf,Type=blockio
>        Lun 4 Path=/dev/sdg,Type=blockio
> 
> We're not using multipathing.
> 
> We've been testing a bunch of configs trying to nail this down, and our current theory is that there's a bug on the Ubuntu target side which causes an iSCSI reset on the initiator side.  When this reset is occurring, we're getting the mutex_panic from what the iscsi initiator is doing with iscsi_handle_reset.  We have some similar systems which work just fine, so we're trying to isolate the differences.  One difference is that we are using 802.1q VLAN tagging on the new setups, and we're not on the old.  We think that might be what's contributing to the reset call.  Regardless, though, the panic on OI is our biggest concern.  

I'm just jumping right in, but I spotted this commit:

http://hg.openindiana.org/illumos-gate/rev/d5c7e34a4e3f

It might be worth you building/installing the latest Illumos gate to see if that addresses the issue.

We're doing reasonably regular Illumos builds, so if you can wait a bit I can ask someone to pop up a new build with this changeset.

Cheers,

Alasdair