Synchronous vs Asynchronous?

[ITKB]

Hiya,

To give you a little insight on our setup:

1 Management Group , 1 Cluster (configured as Multi-Site, with 4x HP-P4300's), and 2 sites (assigned 2 nodes each).

We originally had 2x P4300's with several Network RAID-10 volumes. In preparation for a redundant site at a colo, we ordered 2 more nodes, converted the group/cluster into a Multi-Site configuration, and threw them into the mix. So now we have a Primary and a Secondary site, 2 nodes each. I mostly followed this method of doing so: http://cosonok.blogspot.com/2011/08/converting-existing-hp-p4000-storage.html

My question is...

Is this kind of multi-site setup forced to be of the synchronous model? If this is true, then my understanding is that it is not considered a 'write' until the data is written at both locations. With a smaller bandwidth connection between the sites (say 25MB), will this cause the write speed to the SAN to be incredibly slow?

Is there any way to have this kind of setup, but be async, such as scheduled intervals of Remote Copies? We have come to realize that we will not be able to acquire the kind of bandwidth the documentation recommends (50MB per node, per site), and so I would like to get this back to a configuration that would work with a smaller pipe.

Thanks in advance for any help/advice!

[oikjn · Accepted solution]

asyn is definitely the way you want to go.  You are correct about your assumptions on the writes of your current setup, but the bandwidth might not be your issue if you aren't putting more than 25MB/s through your setup, but LATENCY will be an issue and your total system latency will be the round trip latency between the two sites.

The best thing you should do is REMOVE the two new nodes from the mangement group and create a NEW management group for that 2nd site.  You can then setup a remote snapshot volume between the two groups/sites.  Two management groups are ideal over one group because that gives the two sites complete independance from eachother.... if you have both sites on the same group, you will have to worry about quorum as well and you don't want to have to worry about that during a DR event.

As with any even node installation, you really need to install a FOM for management group (in this case one for each group)...  that FOM needs to be stored on a local HDD (not on the VSA storage itself)...  the specs on the FOM are very modest so it can run on a 5400RPM drive if you want, but try and find a system that is going to have decent up[time for the FOM.

You COULD keep everything in the same managment group and just create two clusters in the group, but this is not a smart idea unless you have a THIRD physical site that you can host a FOM on.

[ITKB]

Thanks for the quick response!

So I can safely remove the 2 nodes from the Second site out of that management group and into another? Will this automatically break the link for the Mutli-Site Cluster and dismantle the site? Any data resyncing?

We intend to make our colo the primary site and would like it to failover to the Secondary site's most recent/complete snapshot in the event of a volume/system failure. How will it understand the failover technique having a FOM at 2 different locations?

Thanks!

[oikjn]

At this point, if you had your four nodes running as one cluster, they are striped across the four nodes and will requore a re-stripe on removal....  that may take a while :)

you should right click on the cluster in CMC and select "remove storage systems"  select both new systems you want to remove and it will let you re-stripe once instead of causing two restripes as you remove both systems.  This is going to take a long time as the data is migrated off, but eventually you will be allowed to remove the nodes from that existing management group and then you can create a new group.

If you are uncomfortable with this process you can always practice if you have a computer that you can install a bunch of VSAs on and try it with the trial installs.  No matter what, I would make sure your backups are good just in case...  I've make this change a dozen times and never had an issue, but Iany time you are working with live company data its always good to cover your bases just in case.

[ITKB]

Alright, will do.

Thanks again! =]