ID #1067

Why do I see reservation conflict messages in the syslog of one node of my NCP 3.x HA cluster?

Nov 20 12:00:01 nodeA RSF-1[2000]: [ID 100000 local0.info] RSF-1 version 3.8.8 OK on nodeA, (hostname nodeA, XXXXXXXXX)
Nov 20 12:02:25 nodeA scsi: [ID 107833 kern.warning] WARNING: /scsi_vhci/disk@g0000000000000000 (sd15):
Nov 20 12:02:25 nodeA        reservation conflict

Reservation conflict messages on one node (node A) of the HA cluster may appear in syslog because PGR3 reservation keys of type WERO (Write Exclusive, Registrants Only) behave as if they are of type EARO (Exclusive Access, Registrants Only) on NCP 3.x, i.e. both read and write access are prohibited if the disk has been reserved by another node.
 
The drive that triggers the reservation conflict message is currently reserved on the other node (node B) where the pool is imported, so the reservation conflict message on node A will be the result of some process attempting to probe the drive on an hourly basis but can’t as it’s been fenced out by SCSI PGR3 mechanism on node B, hence the reservation conflict returned by the OS.

This is not a problem on v4.x as the illumos kernel behaves correctly with regards to PGR3 reservation keys of type WERO, so the probing would be allowed (providing it was a read and not a write) and the OS wouldn’t return a reservation conflict.

Tags: -

Related entries:

Last update: 2013-11-20 11:18
Author: Luis Tortajada
Revision: 1.3

{writeDiggMsgTag} {writeFacebookMsgTag} {writePrintMsgTag} {writeSend2FriendMsgTag} {writePDFTag}
{translationForm}
Please rate this FAQ:

Average rating: 5 (1 Vote)

completely useless 1 2 3 4 5 most valuable

You cannot comment on this entry