Project

General

Profile

Bug #1928

zfs send/receive kills COMSTAR

Added by The Written Word Inc almost 9 years ago. Updated over 8 years ago.

Status:
New
Priority:
Normal
Assignee:
-
Category:
-
Start date:
2011-12-23
Due date:
% Done:

0%

Estimated time:
Difficulty:
Medium
Tags:
needs-triage
Gerrit CR:

Description

We're running COMSTAR on oi_151a with the iSCSI LUNs backed by ZFS volumes. Whenever we try to back up the server, doing a zfs send/receive to the backup server, the iSCSI clients get timeout errors from the COMSTAR server. We set the tunables iscsi_sess_logging, iscsi_login_logging, idm_conn_logging, and idm_sm_logging to 1 while performing the latest zfs send/receive.

I'm including a copy of /tmp/messages with the result of the above tunables set to 1.


Files

messages (1.41 MB) messages The Written Word Inc, 2011-12-23 09:57 PM
lockstat (111 KB) lockstat The Written Word Inc, 2012-02-18 06:41 PM

History

#1

Updated by Dan McDonald over 8 years ago

Have you tried running "lockstat sleep 10" (or some longer snapshot of time) during the time it's hanging? I don't understand the interactions between COMSTAR and zvols as well as I'd like, but I seem to recall that COMSTAR does a lot of direct-zvol access (perhaps more than it should), and you may be seeing lock contention there.

#2

Updated by The Written Word Inc over 8 years ago

Haven't tried that. Will initiate a backup tonight and try. Thanks.

#3

Updated by The Written Word Inc over 8 years ago

Attaching the result of "lockstat sleep 5" when we were performing a zfs send/receive. The iSCSI initiator started failing all clients.

Also available in: Atom PDF