Avatar of sunhux
sunhux
 asked on

Win 2008 R2 cluster quorum disk 'disappeared' temporarily

We had a cluster service 'lost' for several minutes yesterday
(& this is a repeat incident): refer to attached screens.

Q1:
What could be the cause of the issue?

Q2:
Any resolution or workaround for this issue?

Q3:
Currently the heartbeat goes thru Production network.
Will setting up dedicated heartbeat (ie a direct cross-
cable between the 2 member servers of the cluster
resolve this?

Q4:
Or can we tune/tweak the heartbeat interval & the number
of missed heartbeat to make the cluster more resilient
ie to address this issue?  If so, can point me to a link or
provide instructions on how to tune these?

Q5:
Can this be due to differences in the various firmware
(UEFI, versions between the 2 member servers as IBM
found the versions to be different but can't pinpoint that
this issue is due to differences in firmware version of
the member nodes.

I prefer not to attach the Event Viewer logs here as it
contains sensitive info
1.jpg
2.jpg
3.jpg
4.jpg
5.jpg
6.jpg
10.jpg
ClusharedVols.jpg
Clustorage.jpg
Windows OSWindows Server 2008Microsoft SQL Server 2008

Avatar of undefined
Last Comment
Netflo

8/22/2022 - Mon
SOLUTION
Seth Simmons

THIS SOLUTION ONLY AVAILABLE TO MEMBERS.
View this solution by signing up for a free trial.
Members can start a 7-Day free trial and enjoy unlimited access to the platform.
See Pricing Options
Start Free Trial
GET A PERSONALIZED SOLUTION
Ask your own question & get feedback from real experts
Find out why thousands trust the EE community with their toughest problems.
sunhux

ASKER
It's just plain SAN LUNs that's presented to both servers.
We have similar setup at a few other sites' clusters but
did not encounter this issue.

This issue is not persistent, ie it happens every couple
of weeks or months
sunhux

ASKER
I've done  "cluster /prop" & below is the output.  Compared with all other
clusters & only one line (EnabledSharedVolume) is different from other
clusters (except I have one DR cluster which is also 1 but never had
this issue):

Attached are the cluster.log files from both nodes of the cluster
which I've sanitized: look for 2003/12/16-11:50  to  2003/12/16-12:00
date & timings

Listing properties for 'NNXXXPP1VIR03':

T  Cluster              Name                           Value
-- -------------------- ------------------------------ -----------------------
DR NNXXXPP1VIR03        FixQuorum                      0 (0x0)
DR NNXXXPP1VIR03        IgnorePersistentStateOnStartup 0 (0x0)
SR NNXXXPP1VIR03        SharedVolumesRoot              C:\ClusterStorage
D  NNXXXPP1VIR03        AddEvictDelay                  60 (0x3c)
D  NNXXXPP1VIR03        BackupInProgress               0 (0x0)
D  NNXXXPP1VIR03        ClusSvcHangTimeout             60 (0x3c)
D  NNXXXPP1VIR03        ClusSvcRegroupOpeningTimeout   5 (0x5)
D  NNXXXPP1VIR03        ClusSvcRegroupPruningTimeout   5 (0x5)
D  NNXXXPP1VIR03        ClusSvcRegroupStageTimeout     7 (0x7)
D  NNXXXPP1VIR03        ClusSvcRegroupTickInMilliseconds 300 (0x12c)
D  NNXXXPP1VIR03        ClusterGroupWaitDelay          30 (0x1e)
D  NNXXXPP1VIR03        ClusterLogLevel                3 (0x3)
D  NNXXXPP1VIR03        ClusterLogSize                 100 (0x64)
D  NNXXXPP1VIR03        CrossSubnetDelay               1000 (0x3e8)
D  NNXXXPP1VIR03        CrossSubnetThreshold           5 (0x5)
D  NNXXXPP1VIR03        DefaultNetworkRole             2 (0x2)
S  NNXXXPP1VIR03        Description
D  NNXXXPP1VIR03        EnableSharedVolumes            1 (0x1) <==
D  NNXXXPP1VIR03        HangRecoveryAction             3 (0x3)
D  NNXXXPP1VIR03        LogResourceControls            0 (0x0)
D  NNXXXPP1VIR03        PlumbAllCrossSubnetRoutes      0 (0x0)
D  NNXXXPP1VIR03        QuorumArbitrationTimeMax       20 (0x14)
D  NNXXXPP1VIR03        RequestReplyTimeout            60 (0x3c)
D  NNXXXPP1VIR03        RootMemoryReserved             4294967295 (0xffffffff)
D  NNXXXPP1VIR03        SameSubnetDelay                1000 (0x3e8)
D  NNXXXPP1VIR03        SameSubnetThreshold            5 (0x5)
B  NNXXXPP1VIR03        Security Descriptor            01 00 04 80 ... (280 byte
s)
D  NNXXXPP1VIR03        SecurityLevel                  1 (0x1)
M  NNXXXPP1VIR03        SharedVolumeCompatibleFilters
M  NNXXXPP1VIR03        SharedVolumeIncompatibleFilters
D  NNXXXPP1VIR03        ShutdownTimeoutInMinutes       20 (0x14)
D  NNXXXPP1VIR03        WitnessDatabaseWriteTimeout    300 (0x12c)
D  NNXXXPP1VIR03        WitnessRestartInterval         15 (0xf)
DB1clusani.txt
DB2clusani.txt
ASKER CERTIFIED SOLUTION
Netflo

THIS SOLUTION ONLY AVAILABLE TO MEMBERS.
View this solution by signing up for a free trial.
Members can start a 7-Day free trial and enjoy unlimited access to the platform.
See Pricing Options
Start Free Trial
GET A PERSONALIZED SOLUTION
Ask your own question & get feedback from real experts
Find out why thousands trust the EE community with their toughest problems.
sunhux

ASKER
> are your NIC configured to go to sleep - this is a default setting under the
> NIC power management setting, turn that off.
No, certainly it's not, else I would have got Tivoli alert for ping
timeouts to both DB servers'  teamed NICs' IP addr.

>  validate your cluster as that may highlight some issues
We have run this & only a couple of warnings with the rest
Successful.  I can post it this Mon/Tue after sanitizing the
outputs.  For sure, it's the same result as when we first
install the cluster & if any of the error/warnings are serious,
cluster installation would not have been allowed by Windows
in the 1st place, right?
All of life is about relationships, and EE has made a viirtual community a real community. It lifts everyone's boat
William Peck
SOLUTION
Netflo

THIS SOLUTION ONLY AVAILABLE TO MEMBERS.
View this solution by signing up for a free trial.
Members can start a 7-Day free trial and enjoy unlimited access to the platform.
See Pricing Options
Start Free Trial
⚡ FREE TRIAL OFFER
Try out a week of full access for free.
Find out why thousands trust the EE community with their toughest problems.
sunhux

ASKER
Yes, all components have been rebooted.

Well, I'm not sure if it's really necessary to apply the hotfixes as the
other sites' cluster uses exactly the same Win 2008 R2 with same
patch level.

Only difference is the hardware type & firmware revision:
this site uses a slightly newer version of IBM x3850 hardware
compared to the others sites that are stable.

Is there anything we can analyse from the logs to pin-point
the root cause so that we can come up with a sure-fire
resolution?
SOLUTION
Netflo

THIS SOLUTION ONLY AVAILABLE TO MEMBERS.
View this solution by signing up for a free trial.
Members can start a 7-Day free trial and enjoy unlimited access to the platform.
See Pricing Options
Start Free Trial
⚡ FREE TRIAL OFFER
Try out a week of full access for free.
Find out why thousands trust the EE community with their toughest problems.