We will be doing journaling to a new EV system from exchange 2007.
We will be having CA and DA installed on separate machines, but will using the same SQL server as EV.
We will have a DR site, that will replicate all data on SAN level.
The DR site will contain 1 EV server and one SQL server.
Please comment the setup and the questions i have.
This is the design we were thinking of using,
One Microsoft 2008 Cluster running Enteprise Vault. Two node Active- Passive.
The Cluster will only be doing journaling from 1 or 2 Journaling mailboxes in a CCR cluster. (No regualar Mailbox archiving)
Expected data growth will be 1,5 - 2.0 Tb a year.
OS 2008 Ent edition
OS and EV installed on local disks,
Index and MSMQ will be on SAN, FC disks.
Vault store, CIFS WORM. on SAN SATA disks
HW: 2 quad core cpu, 16 Gb mem per server.
One EV Discovery Accelerator machine on a VM
One EV Compliance Accelerator machine on a VM
OS 2008 sp2 x64
HW: 4 vCPU
Mem: 4 Gb
Expected DA searches will be 1 a month. so it wil not be heaviy used.
One SQL server that will host DB for EV, DA and CA.
OS, 2008 sp2 x64
HW, 2 quad core cpu, 16 Gb mem
Site2 (will be used as DR site)
Data replicated on SAN level
1 EV server joined to the same EV site, we will be using building blocks for failover.
Same specs as Ev servers in site1
1 SQL server.
Same specs as the SQL in site1
Comments of the setup ?
1. If site1 fails and we have to to a failover to the DR site (site2), how do we do that in the best way?
The EV server we will be using the USL (building blocks failover)
But what do we do with the SQL? the SQL in the DR site will have a diffret hostname but the same data. Do we have to manuall reconfigure what SQL the EV site should point to ? What is the best way for this?
2. What about specs for the Disks sizes ?
We were thinking of using several 300 Gb Luns for the Index, and one big 8 Tb CIFS WORM share for the Vault Storage. comments? Should i split up te CIFS share in smaller Luns? how big if so? benifits of that ?
3. If SQL and or EV servers in site1 fails and we have to do a "failover" to the DR site, what must be done to point the DA and CA server to the new SQL name in site2?
4, Recomended luns and lun sizes for the DA and CA servers ? They will not be heavily used.
5. Does EV support SQL Server database mirroring for automatic fail over? could not find anything of that in the Docs.
I probbably have some more questions for you guys later :D
Get your MSMQ also on local disk. I believe SAN is not supported/too slow
DA/CA not supported on VM. Will work, but Symantec recommends these running on physical machines.
You sql is more than capable of handling ev.
1 - you will have to change sql-server that ev approaches. You could (but test!!!) see if you can use an alias. We use this in our environment, as we do not have the default port used by SQL. You can use on the DR-EV-server :
start, run, cliconfg
give OLD sql server as aliasname, and point to the new server as where it should go. Might work!!! again, test!
2 - 300GB for index is more than sufficient.
I assume (as you use WORM) you will not expire from journal. I'd personally go for smaller cifs luns, and use ev-partitions to keep the data manageable. example. 2tb a year is approx 170 GB a month. make partitions of 200GB, and configure partition roll-over using diskspace as value.
This means if you need to backup worms, once a partition has rolled over, it won't change anymore (assuming you do not expire!), and youneed not make a frequent backup of that partition (no worms expert though)
3 - use cliconfg as easiest workaround. There is a document (http://seer.entsupport.symantec.com/docs/340611.htm) that describes moving databases to other sql server.
4 - DA/CA (no experience in CA) DA searches can produce massive results. Probably best having a 500GB disk for results.
The MSQM will be clustred and held on a fast disks on the SAN, no problem there.
DA and CA is supported on VM, there is a whole section of virtualisation for DA in the best practisce guide for DA.
regarding using 200 gb luns, there will be alot of luns to manage. that would be 10 luns a year with 1 partition on each.
Seems alot to me.
We will not be expire items, or maybe we will after for example 10 years. then we will remove the lock on thoose luns in the san and let ev expire.
Backup will be snapshotting and replication in the SAN