System Freeze behavior has changed between VCS version 4.x and 5.x
Article ID: 100004449
Updated On:
Description
Error Message
Resolution
The behavior of freezing the system in a VCS cluster has changed dramatically from version 4 to version 5. In version 4, "freezing a system" meant that none of the Service Groups could fail over from that system. In version 5 and above, however, it means that Service Groups can still fault, and fail-over to other systems from a frozen system. Further, a frozen system in version 5 and above means that Service Groups cannot come online while that system is frozen.
| VERSION | Behavior |
| 4.1 | SGs cannot fail over to other system when system frozen |
| 5.0 | SGs can still fail over |
| 5.0 MP3 | SGs can still fail over |
| 5.1 | SGs can still fail over |
Here is an example of the change in behavior illustrated on a linux RH5 system running VCS 5.0 MP3
[gfellin@dex ~]# su -
[root@dex ~]# rpm -qa | grep VRTSvcs-5
VRTSvcs-5.0.30.00-MP3_RHEL5
[root@dex ~]#
[root@dex ~]# hastatus -sum
-- SYSTEM STATE
-- System State Frozen
A dex RUNNING 0
A polly RUNNING 0
-- GROUP STATE
-- Group System Probed AutoDisabled State
B websg dex Y N ONLINE
B websg polly Y N OFFLINE
[root@dex ~]# haconf -makerw
[root@dex ~]# hasys -freeze -persistent dex
[root@dex ~]# haconf -dump -makero
Fault a resource on the frozen system.
[root@dex ~]# umount /webvol
[root@dex ~]# hastatus
attempting to connect....
attempting to connect....connected
group resource system message
--------------- -------------------- -------------------- --------------------
dex RUNNING
polly RUNNING
VCShmg dex OFFLINE
VCShmg polly OFFLINE
-------------------------------------------------------------------------
websg dex ONLINE
websg polly OFFLINE
VCShm dex ONLINE
VCShm polly ONLINE
webmount dex ONLINE
-------------------------------------------------------------------------
webmount polly OFFLINE
webvolume dex ONLINE
webvolume polly OFFLINE
webdg dex ONLINE
webdg polly OFFLINE
-------------------------------------------------------------------------After the fault is detected, the entire SG fails over to the other node (polly)
websg dex PARTIAL *FAULTED*
webmount dex *FAULTED*
websg dex STOPPING PARTIAL *FAULTED*
webvolume dex WAITING FOR OFFLINE
webvolume dex OFFLINE
-------------------------------------------------------------------------
webdg dex WAITING FOR OFFLINE
webdg dex OFFLINE
websg dex *FAULTED* OFFLINE
webmount polly WAITING FOR CHILDREN ONLINE
webvolume polly WAITING FOR CHILDREN ONLINE
-------------------------------------------------------------------------
webdg polly WAITING FOR ONLINE
websg polly STARTING OFFLINE
webdg polly ONLINE
websg polly STARTING PARTIAL
webvolume polly WAITING FOR ONLINE
-------------------------------------------------------------------------
webvolume polly ONLINE
webmount polly WAITING FOR ONLINE
webmount polly ONLINE
websg polly ONLINE
With node "dex" still frozen, clear the fault, and then take the resource offline on the other node. Since "dex" is still frozen, then entire SG faults, but does not fail over, since it has no where else to go in this two node cluster:
[root@dex ~]#
[root@dex ~]# hares -clear webmount dex
[root@dex ~]# ssh polly "umount /webvol"
[root@dex ~]# hastatus
attempting to connect....
attempting to connect....connected
group resource system message
--------------- -------------------- -------------------- --------------------
dex RUNNING
polly RUNNING
VCShmg dex OFFLINE
VCShmg polly OFFLINE
-------------------------------------------------------------------------
websg dex OFFLINE
websg polly ONLINE
VCShm dex ONLINE
VCShm polly ONLINE
webmount dex OFFLINE
-------------------------------------------------------------------------
webmount polly ONLINE
webvolume dex OFFLINE
webvolume polly ONLINE
webdg dex OFFLINE
webdg polly ONLINE
-------------------------------------------------------------------------After the fault is detected, the entire SG is faulted, but does not fail over:
websg polly PARTIAL *FAULTED*
webmount polly *FAULTED*
websg polly STOPPING PARTIAL *FAULTED*
webvolume polly WAITING FOR OFFLINE
webvolume polly OFFLINE
-------------------------------------------------------------------------
webdg polly WAITING FOR OFFLINE
webdg polly OFFLINE
websg polly *FAULTED* OFFLINE
Issue/Introduction
Freezing a system Freeze a system to prevent the service groups that it hosts from failing over to another system.
In v5.x VCS and above:
Freezing a system Freeze a system to prevent service groups from coming online on the system.
However, they still can fail over, or the entire service group can fault, if a critical resource goes offline while frozen.
Additional Information
