11-03-2010 08:47 PM
We are using this agent for the replication of LUNS in the cluster environment.
The version of the documentation I have is Version 5.1.
It states in the document the following on page 7
The agent supports configuring EMC MirrorView in the synchronous or
asynchronous modes. In asynchronous mode, you can replicate either individual
LUNs or replicate consistency groups. MirrorView can also replicate LUNs or
metaLUNs. In synchronous mode, you cannot replicate consistency groups.
I interpret this to mean that if we use synchronous replication we cannot use consistency groups.
However I have been led to believe that it is in fact supported by other people.
Can someone give me a clear and definite statement on this matter as to whether it is supported or not.
Regards bob mitchell;
Solved! Go to Solution.
11-15-2010 11:23 PM
Hi,
I think you've taken everything into account. I hope that the initial issue about luns & consistency groups isn't a problem though, dont know to much about the EMC internals, but I suspect that since you wont be having to many things to replicate (like logs and dbf's, only indexes) you should get away with just a lun(s).
Good luck :)
11-07-2010 09:09 PM
Hi,
The latest guide (5.1 Solaris 10 as an example) states
In synchronous mode, you can replicate individual LUNs but cannot replicate consistency groups.
Who are "other people"?
R
11-09-2010 07:25 PM
At first one of our staff interpreted this incorrectly believing it referred to mirrorview only and not the AGENT software.
To resolve an argument i logged a call to Symantec Support who also at first said it was supported.
I then referred them to probably the same section of the agent installation and config guide as you have read and they recanred and aplogised for misleading me and confirmed that it is not supported.
We have a company on site who has an EX Symantec person in their team who is also not accepting it yet even though it is black and white.
So I was seeking confirmation from other places!!
Regards Bob mitchell
11-09-2010 08:43 PM
Hi,
Ok, so its still a bit unclear. Unfortunately I've only ever used SRDF, so I can't comment on real world situation.
One thing I would like to ask though is, do you not think it would be better to run async? How dispersed are the two sites, and how big is your bandwidth between the sites, and how reliable is it? From a VVR point of view, I usually run async for all installations, and if bandwidth permits, synchronous override (switches from sync to async in if bandwidth becomes unavaible), never sync.
Just a thought.
R
11-14-2010 05:01 PM
We have a 4 GB/sec links between ou sites which are around 80 KM apart.
With the correct Buffer Credit allocations I do not think we would have a problem.
I designed Synchronous because we are primarily relying on Replication for our INDEXES as the actual Vault Store is on Centera which is replicated VIS IP.
We are not planning to backup the INDEX stores but rely on rebuilding indexes if we get corruption or index loss of some sort. (we are expecting very large amounts of index data as we are planning on over 150,000 user accounts in Exchange all with archiving enabled.
I did not want to introduce any window of opportunity for data to be in a non-synchronised state if a failover occurs. Hence I ruled out Asynch.
Would you think this is a reasonable strategy - i must admit I am a little concerned with the possible impact of synchronous replication on the archiving rate but everything I have been told/read indicates that this rate is generally bottlenecked by CPU of the indexing server (although I am not convinced).
Regards Bob Mitchell
11-14-2010 07:29 PM
Hello Bob,
to my experience, whether to use sync or async is always a matter if business decision. As you explained the reasons, sounds fair to use sync mode .. however just thinking, what if you use sync mode & network link has some unplanned outage ? I believe there would be a redundant plan but in case of issues with network link, sync mode will not even complete writes in primary (talking from VVR prospect) ..
just hoping if this point of view has been considered..
Gaurav
11-15-2010 04:50 PM
our fabric has no single point of failure so a "double" failure would need to occur and if this was to happen we would have a cluster failover to a "integral" copy - which is my design.
Under this scenario we could then run in non redundant mode and orchestrate a sync manually in order to re-establish "normal" operations.
thanks for your feedback - i value any comment/suggestions/questions - it makes you think things thru.
Regards
11-15-2010 11:23 PM
Hi,
I think you've taken everything into account. I hope that the initial issue about luns & consistency groups isn't a problem though, dont know to much about the EMC internals, but I suspect that since you wont be having to many things to replicate (like logs and dbf's, only indexes) you should get away with just a lun(s).
Good luck :)