This post is more than 5 years old
21 Posts
0
5233
Vmware SRM and consistency groups
Hi guys,
I have a client who is using a cx3-10 and are replicating using mirrorview/a
They have 8 luns in a single consistency group which are replicating to the DR site.
The exchange vm's vmdks placement is, the system vmdk on one lun, logs on a second and DB on a third.
I was wondering when you replicate the luns would the consistency group understand the relationship between the exchange DB and logs?
If not would it be better to place the logs and DB on the same LUN?
If anyone can point to some docs about this, or has any ideas it would be much appreciated.
Thanks
Dougie
faisalchoudry
53 Posts
0
July 15th, 2010 09:00
Firstly regarding the 8 LUNS within the Consistency group.
You don't need to have all 8 LUNS in the same consistency group, if these are just separate virtual machines/VMFS datastores. I would just keep a consistency group relevant to the Exchange system you are trying to protect.
(You can have a consistency group per Exchange storage group)
For an Exchange server on a VM, you'll have the following to protect and replicate:
1)The VMware datastore that holds the bootfile (VMDK) for the Exchange operating system.
2)An Exchange Database LUN(s) (lets assume 1 database per Exchange storage group)
3)the Exchange Log Lun
so as an example you could put the above 3 LUNS into a Consistency group.
Actually I just noticed you said you've an Exchange Cluster so okay I see why you have so many LUNS in the consistency group.
I take it they are all related to the Exchange Cluster and Exchange ?
As you are only using MirrorView/A and no scripting or other tools to replicate, then your replica on the other site will be a "Crash Consistent" copy of Exchange. If you recover (fail over/promote), you may have to run some tasks on the Exchange system you are recovering at the other site in order to get it up and running.
See if the docs below help and let me know.
You can also try a test recovery. if you have SRM you can use the test option, then have a look at the recoverd vms(s).
Or manually you can snapshot the secondary LUNs (of each mirror pair) and mount those snapshots onto the standyby server at the other side, bring up the Exchange system, and have a look. Make sure the network you bring the vms into, within ESX does, not conflict with the actual network for the primary machines as you dont want the test copies you are bringing on-line to interfere with the production Exchange system.
Have a read of :
1)MirrorView Knowledge Book FLARE 29 (in Powerlink)
faisalchoudry
53 Posts
0
July 15th, 2010 06:00
Placing the Exchange Logs and and Exchange DBs on separate LUNS is actually best practice, for resilience.
There is a Best Practice paper within Powerlink, on Exchange and Clariion configuration.
I havnet copied a paper into here as you havent mentioned the version of Exchange you have - if you have any trouble finding a paper let me know some more detail and I'll have a look in Powerlink.
The Consistency Groups within MirrorView /A.
When you add multiple LUNs into a consistenct group, it's normally because those LUNs are assocaited to the same application so,
with replication, you wouldnt want replication to continue for one LUN but not the other, so in a consistency group, for whatever reason,if there is a problem with replicating 1 LUN, then a Fault condition appears on the entire Consistency group and all the LUNs go into a factured state - which is good, you want to know there is a problem.
Also you manage all the associated LUNs via the consistency group object, so for example if you had to failover (promotion), you do that for the Consistency group, not each indivudual LUN, and then promotion takes place for all the LUNS - what you would want in a disaster situation - minimise the number of steps.
For the older versions for SRM (pre ver 4) you also required consistency groups, as otherwise SRM would not pick up any LUNs to protect, even if just single LUNs. Not the case now with SRM 4.
why do you have 8 LUNS in a consistency group though ? are they all associated ?
one pain can be if one LUN stops replicating witin the consistency group, they all stop, so best to have only LUNs associated to the same application within a consistency group.
douglas_carson2
21 Posts
0
July 15th, 2010 07:00
Hi Faisal,
There are 8 LUNs in the consistency group as there are 8 VMFS volumes being presented to the ESX cluster.
The client has spread their vms vmdks between different VMFS volumes. As you said this is best practise to spread your vmdks on different LUNs and RAID levels. So in a way for all the VMs to be replicated across to the DR site all the VMFS volumes dpend on each other.
I guess my question was more around the state of exchange at the DR site. I was trying to figure out when i perform a test recovery will the exchange logs and database be in a consistent state to bring up exchange.
Thanks
Douglas
douglas_carson2
21 Posts
0
July 15th, 2010 11:00
Thats Great thanks Faisal.
I appreciate the help.
Best Regards
Dougie