Exchange P4500G2 with P4530 - Questions

[Exploitation67]

Hello,

I will soonly exchange systems for my 4 nodes P4500G2 multisite cluster with new Storevirtual P4530 and I have few questions; all my volumes are in RAIN10 and I've got a FOM in a 3 third site.

- First I have to exchange 2 old P4500G2 (currently in production) with 2 new P4530, one in Site A and one in site B. Can I exchange the 2 systems at the same time? If one site is lost, will the other site work properly to maintain access to the volumes?

- What is the best ratio to set in the MG Local Bandwith properties? Is the maximum a really good option?

When I set it manually to 10MB/s, sometimes some volumes go down (on windows server side - with HP DSM for MPIO installed); all nodes are connected with ALB teaming with 2*1Gbits links. For the moment, the rate is set to 2MB/s and all works fine but I think that the time to copy the data (because my cluster is nearly full I've got approximatively 5TB to copy from one P4500G2 to the new P4530) will be too long and I don't want to wait too much time.

Have you got some best practices?

Many Thanks for your help.

[oikjn]

nobody can tell you the correct speed...  you will just have to adjust and see where the performance impact on your system is acceptible and use that.  When I have exchanged nodes, I've kept it <5MB and just let the thing run for a long time since I was not in a rush and it did not affect our performance.

You can exchange multiple nodes at the same time and I believe it is perferrable since it takes much less time overall.

If you are in a rush, the only thing you can do is put the bandwidth up to full, but as you see that will really affect your access to your LUNs.  The only way you could do full bandwidth would be if you had a maintenance window where you could shut everything accessing the system off so all resources can be dedicated to the migration.

Why the time rush?  If this is for returning the old units you really just need to talk with them to have that time extended since it sounds like the time you got from them to make the return is unrealistic.

[Exploitation67]

Hello,

Thanks for your reply, but I'm just confused about the operations that I have to do :

1)  Install the 2 new nodes and integrate them into the existing MG.

2)  I have to edit the existing cluster but I have to choose which option : Exchange Storage Systems or Add Storage Systems and in this box remove the old systems?

The swap storage systems option is, I think, to "exchange"all the systems in one cluster.

Another question is : Have you got some mathematics to estimate the time of the restripe operation?

Thanks for your help.

[Exploitation67 · Accepted solution]

....or can I use the "swap storage Systems" and only choose systems that I have to "exchange"?

[oikjn]

The one you want is "Swap".  The exchange is used for replacing missing nodes.  I thought you could swap more than one at a time, but reading the help from CMC is seems to say you have to do it one at a time.

[HPstorageTom]

You can swap multiple systems at a time. And if you want to do a complete cluster swap, then it is the best to swap all cluster nodes in one go (avoids multiple restriping operations).

[Exploitation67]

Hello,

Thanks all for your reply, the swap option was the solution.

I "swapped" 2 storage nodes in one time and the restriping operation complete in about 8 days at 8MB/s maximum with no impact on production.

I just add 2 more nodes to my cluster and the restriping is running again.

During the restriping operation, can I delete a volume or the restriping of this volume must to be completed?

[oikjn]

I'm 99% sure you can delete a volume while it is being restriped without an issue.  I can't remember last time I tried BOTH at the same time, but if you are trying to delete a volume I don't see why you shouldn't try to delete it... worst case CMC just rejects the operation.  HP did a pretty good job of not letting you do something that will really break the SAN