We've built a SAN based on NexentaStor 3.0, and have generally had a good 
experience with the flexibility and performance of ZFS and the relatively new 
COMSTAR kernel mode iSCSI story.

However, we are experiencing a frustrating and thus far terminal robustness 
issue which is keeping the box out of production: whenever we initiate iSCSI 
from our Hyper-V cluster nodes and turn on deduplication and compression (and 
sometimes even when we turn both of them off), and we hammer high bandwidth 
serial writes against the COMSTAR iSCSI targets, the OpenSolaris box tends to 
trigger timeouts and other iSCSI failures in the Hyper-V iSCSI initiator. In 
the case of our Hyper-V cluster, this causes our Cluster Shared Volumes to 
fail, which in turn causes all of our VMs to reset.

This thread describes the issues as seen on the initiator side: 
http://social.technet.microsoft.com/Forums/en/windowsserver2008r2general/thread/4b8f659f-2380-48e2-960a-97048bd5cb1c.

Only rarely has a reset of the OpenSolaris improved the issue in any way 
whatsoever, which leads me to believe it is generally healthy. Increasing the 
timeout value on the Hyper-V R2 side as described in the above thread is not 
solving the problem.

So here's my initial question: where are the log entries that I should be 
reviewing and posting on the OpenSolaris side with respect to COMSTAR iSCSI. 
First and foremost, I want to diagnose whether the OpenSolaris box is unhappy 
when these failures are occuring.

For reference, the selection of initiator errors in the Windows System Log are:
* Warning iScsiPrt 129
* Error iScsiPrt 9 - target did not respond in time for SCSI request. The CDB 
is given in the dump data.
* Error iScsiPrt 39 0 initiator sent a task management command to reset the 
target..

A full reset of the initiator generally rectifies the issue. Occasionally, the 
problem self corrects after various timeout and reset times expire on the 
Windows side.

Please advise.

TIA, Andrew Jones
-- 
This message posted from opensolaris.org
_______________________________________________
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org

Reply via email to