Vault High Availability

[vtvv11]

I should set up Enterprise Vault 9 for Exchange 2010 with DAGs. No problem with that.

I'd be interested in getting opinions for basic architecture, when the requirement is that the archiving system is doubled into two data centers. That means also archived data should reside in two separate data centers.

As far as I understand, Vault does not replicate or mirror data in any way, I'm wondering what would be a reasonable way to set this up.

[WiTSend]

Generally you can either use a hardware platform native replication process, or you'll need to set up some like VVR to replicate the Vault Stores, Indexes and SQL databases to the secondary datacenter. For the EV servers: You can either use SFW/HA w/GCO and set up a geographic cluster, or you can rely on EV's USL process to provide the failover to the other site. If you are using virtualized (VMWare) servers you can also use VM tools to provide the failover of the computers.

[SolarP]

Good question as I have a similar scenario coming up for a customer.

Firstly can you explain SFW/HA w/GCO.... too many TLAs there ;)

Assuming you have a resilient SQL implementation and Vault stores and indexes are being replicated to a DR site, I would like to know what people are putting out there for EV 9 for local resilience and DR.

My understanding is that if you use clustering (either MS or Veritas) to provide local HA, you can not use USL and building blocks in the DR site to single servers. You have to have the servers configured within a cluster before this can happen. Is this still correct

Alternatively are people using EV within a VMware cluster to provide HA and DR aspects of the solution.

As has been said before this aspect realy does need sorting wth a proper DR options document.

Thanks

SP

[elbutre]

hi

our consultant has just upgraded us from 7.5 to 9.02 and moved it to a new virtualised setup

all is working well and we are now in the process of setting up the DR server.

i am intending on copying the data using robocopy scripts from our production server to our DR server.

i have got SQL mirroring setup (sql 2008 r2 standard with witness server for automatic failover)

is there any documentation whatsover on the symantec site that describes how you setup a DR server not using veritas or ms clustering? the only thing i can see is the chapter about failover in a building block scenario but it does not describe at all best way to configure the DR server.

the EV installation on the DR server has been done but i am wondering whether i should configure things like index locations, shopping locations, on the DR server just like it has been done on the production server or will this need to be done during a DR scenario?

we only have about 100GB of data so WAN replication should work quite well.

many thanks for your help

[wandarah]

Need more info on your DR strategy. 

It could be you can use building blocks, and add your DR server to the EV site as a passive machine. In the event of failure, run the USL tool. This would require that your index locations, etc, are accessible in the same manner (i.e mount point, etc) from the DR machine as well on the same path, yes.

That all depends on your environment, of course.

[elbutre]

my setup is as follows

office has 1 active evault v9.0.2 setup

DR site has a passive evault setup

servers are identical (installation paths and index, shopping service, cache and partition locations)

robocopy scripts copy the index, shopping, cache and vault partitions from the active evault server to the passive one. it has copied 30 out of 100gb overnight so it will be another 2 days  before the DR server is fully up to date, after that point i will run incremental robocoy scripts to keep the DR server up to date.

SQL database is installed locally on each evault server and have setup sql mirroring using a sql witness server.

i understand that in the case of a DR scenario need to run the USL tool. will probalby also only run the evault in read only mode in case need to invoke DR site.

is there any symantec documentation that describes how you set this up, i can only see documentation if you use windows or veritas clustering but not if you have an active / passive setup. or perhaps there is but i have not been able to find it.

[GertjanA]

Hello

The documentation is not very clear, but it is logical to set up.

What you need to (have done) do is the following:

Server1 has alias EV001, server2 has alias EV002

When you do a DR, you set EV in backupmode. You then make sure (!!) all data and sql is replicated/in sync.

This depends on your archiving schedule obviously.

Then you switch the aliasses. (EV001 points to server2, EV002 points to server1), stop (and reverse if needed) replication. Connect to server2, run USL. This should add missing services, tasks etc.

Verify all is running on server2 (now EV001), users can connect using Archive Explorer etc. Set EV in normalmode, off you go. When DR situation is over, replicate data back to server1 (which is still EV002 alias), perform same actions.

GJ

[elbutre]

thanks for that, i have got the DNS aliases setup as you suggested.

i only archive once per month currently so doing overnight evault incremental data replication is good enough for our organisation (sql is mirrored realtime). we only have 25 users so not much happening on the archive side of things

once data replication has finished i will try and do a failover and document the steps

This forum is great for claryfing these steps for someone like me

thanks