Is clustering mode for a PV220s "sharing"?

Hi,

Thanks for the post.

Hmm, my idea doesn't seem so good any more now if i understand the situation correctly. :smileysad:

The new server will be running w2k3 R2 however the existig one is a bog standard W2k server. I suspect that the best course of action would then be to change the current configuration to split bus mode to allow each server/host to use its' respective half of the array.

Would the change be a destructive one in going from joined bus to split bus? I suspect it may.....

Thanks,

Pop

You can only share if you set up a Microsoft Cluster. This will require Windows 2000 Advanced server or Windows 2003 Enterprise (32-bit or x64). If you try to do this without a cluster, you'll end up with data corruption.

You'll need 2 identical raid controllers; one per server.

You'll lose the last slot. This will be completely useless. Also, depending on your current firmware/setup you may need to scratch your raid setup completely, reconfigure the raid controllers to cluster mode and then create the cluster setup.

You'll need a mirror for the Quorum, so you'll be left with 11 drives at most for cluster shared storage space.

Also, in a Microsoft cluster, only 1 server/node at a time can read & write to and from the drive in question, so if you want 2 servers to be able to fully read and write to the same disk/partition at the same time, you'll need to give the disk to 1 server (like it sounds you already have) and then create a share and let the other server map a drive letter to it (going over your network).

That depends on how your current setup is. If you go split mode, the PV220S is cut in half; one management module gets the 7 drives on that side of the enclosure and the other module gets the other 7 drives. So if you have a raid setup that crosses this invisible cut-off point, you'll lose the config.

Interesting. Is it really as simple as saying that once the switch is flipped to split mode, the first 7 disks are recognized to the first EMM while the remainder attach to the second?

 

Currently, there is a R10 container spanning the first 6 disks so under this assumption, they should remain intact. The remaining 8 disks which currently form a R5 container (of which one disk is located physically among the first 6) will be wiped because the container is invalid.

 

On second thoughts, would that second R5 container that's among the remaining 8 disks not just become degraded since the 6th disk is now attached to the first EMM?

 

Hope i haven't lost you so far :smileyhappy: and thanks for your input.

 

Regards

 

Alex

Thanks for the post. A little thought and it seems that this should be the case too.

With regards to the second container that has a drive across the boundary, taking an educated guess i assumed that it's possible the container would be degraded as opposed to fail entirely. I'd appreciate your opinion if that is indeed the case.

Thanks,

Edit: Another thought, but if i just removed the lone drive from the second container so as to remove any cross-over, then flipped the switch and rebooted, wouldn't both containers now remain intact? I think this may work.

IT does require a powercycle of the enclosure (and the server), but it should be that simple.

So:
- make sure to have a tested backup of your data just in case
- power off the server
- power off the PV220S
- change the switch to split mode
- power up the enclosure
- power up the server
- verify in the raid controller that your virtual drive is still recognized

For anyone in the same boat, i'd thought that i would update this thread to provide a brief explanation of how the process went.

 

First off, the aim was to split the array's accordingly so that the logical partition of the whole array was created before changing the bus mode from joint to split. This was necessary so that the array wouldn't return degraded/failed later on.

 

In doing so, i re-configured the RAID containers (two container array: a Raid 5 spanning 6 disks and a Raid 5 spanning 8 disks) to break the second container since the 7th disk was past the physical boundary of the split bus mode and cleared the orphaned data on it to allow it to be joined as a hot spare for the first RAID 5 container.

 

The second container now returns degraded as there is a missing disk from the array as expected. The container still works, but some reconfiguring would be needed to return this to an optimal state. I accepted that i didn't want to spend time going through the steps to restore this container as the data on it was irrelevant and so chose to press ahead with setting the array as a split bus mode.

 

When i powered off and flipped the switch turning the PV from joint bus to split bus, the first container was detected via the one EMM as anticipated, however the remaining config data needed to be cleared up as the second container was no longer visible to this system. This didn't take more than a few minutes and the system accepted the new array with all the data on it intact after the split.

 

The second container (which was by now degraded) had trouble coming on-line as expected so after reading the configuration from the disks as opposed to loading the config from the array, i re-initialised the remaining 7 disks and created a new RAID 5 container connected to the second EMM. This took a while to create, scrub and carry out the necessary integrity checks, but it was ultimately an accepted configuration and returns as optimal now.

 

There was one hiccup however and this was down to a small mistake on my behalf. On separating the arrays in the PV and connecting separate PERCs to the two EMM's, the new server was incorrectly connected to Channel A of the external ports when it should have been connected to Channel B - the internal RAID 1 drives were using the internal ports for Channel A and so this conflict crashed the system. It didn't cause any trouble afterwards however and configuration was easy thereafter.

 

All in all, yes it is possible to split the bus on-the-fly and you will retain all your original data as long as you split the array's beforehand so that the containers do not go beyond the physical boundary when split bus mode is enabled.

 

Regards