W2K8 3PAR LUNs in Maintenance Mode during InFormOS upgrade ?

[eve]

Hi, we soon will be upgrading our 3PAR to a newer InFormOS version.

HP 3PAR has told us that the LUNs we present from our 3PAR to our Windows 2008 Cluster,
need to be put in maintenance mode as this will prevent a server node-failover.
According to HP 3PAR maintenance mode ensures we have no outage during the upgrade.

I might be wrong
But if I am correct, LUNs in Maintenance mode are not available.

So my databases are basically offline during the InFormOS upgrade ?


Hope that somebody can shine a light on this ?

[Richard Siemers]

I am curious about this as well... I brought this information to my windows team and they looked at me like they never heard of maintenance mode before... simply because they hadn't. In the past we enjoyed hot upgrades to the 3PAR without much of any effort on their part.

They reported back that maintenance mode has to be enabled on a lun by lun basis with their admin tools as well. This is road block for our upgrade because we have 20+ clusters and 150+ luns that would need this set.

[eve]

I have done some digging and if I get it all correctly (English is not my native language)
it all depends on if the LUNs are Cluster Shared Volumes or Non-CSV

* if CSV LUNs are put into Maintenance mode, then the data is not available
* if Non-CSV LUNs are put into Maintenance mode, data remains available.

http://technet.microsoft.com/en-us/libr ... 72587.aspx
http://technet.microsoft.com/en-us/libr ... 59255.aspx




Maybe a Windows expert can shine a light on this ?

[Richard Siemers]

According to wikipedia:

Cluster Shared Volumes (CSV) is a feature of Failover Clustering first introduced in Windows Server 2008 R2 for use with the Hyper-V role. A Cluster Shared Volume is a shared disk containing an NTFS volume that is made accessible for read and write operations by all nodes within a Windows Server Failover Cluster.

Maintenance mode is a mechanism provided through cluster.exe and the Failover Cluster API that places the specified resource in a mode that will disable health checking. After maintenance mode is enabled for a resource, Resource Monitor will ignore health check calls on the resource even though the resource is left in online mode.

Do we "really" need to put them in maintenance mode during an upgrade? What is new in 2008 that requires this where 2003 does not? Why isn't multipathing good enough anymore?

[jaishan2002]

I never heard LUNs in Maintenance mode in Windows 2008

Recently we have upgraded our T400 storage boxes, during that time most of the windows 2008 clusters resourses were failied over to another node & few were rebooted.

When we have worked with 3PAR on this issue & they do not find any cause at storage end & they suspected may be some MPIO issue at host end.

I really do not find what is the reason for this.

Thank You
Jayashankar

[macknife]

We upgraded to 3.1.1 last year, but we never had to put the LUNs into Maintenance.

If I remember correctly, only half of the 3PAR controller nodes are updated at a time and rebooted, so assuming hosts are connected across node-pairs, online disk access is maintained throughout the update process. The 3PAR update process does not disrupt I/O processing unlike EVA updates, which (stupidly) reboots both controller nodes at the same time.

[jaishan2002]

Your absoultly right.

I am just wondering what would be the cause to windows 2008 Cluster resources failed over another node and few servers reboot since there were no proper logs avilable at host and Storage end during InForm OS upgrade.

Thank You
Jayashankar

[eve]

Hi All,

I now have a confirmation from HP 3PAR that indeed it is required to put 3PAR LUNs presented to
W2K8 cluster in maintenance mode, but for non-CSV (Cluster Shared Volumes) it has no impact



CSV are only used in HyperV environments.
If CSV are put in Maintenance Mode, they are offline for HyperV

[wkdpanda]

We have run into this before with our upgrades.

I don't know about "maintenance mode", but there was documentation about dealing with the luns. When a lun path failed (because it was offline during the reboot), that failure could lead to cluster failovers/failures.

-Andy