GroupWise on SANs
Novell Cool Solutions: Tip
By Tim Heywood
Digg This -
Posted: 25 Apr 2006
A Forum reader described the following situation:
"I've been tasked to come up with a solution to provide a highly available web based email system to around 4000 users for my organization. Highly available to my CIO means clustering from our primary site to our business continuity/disaster recovery site. If we lose an entire site, our mail system would fail over to the remaining site and keep running.
We currently have an IBM DS4300 SAN at our primary site. If I put another DS4300 SAN at our BC/DR site and purchase Remote Mirroring and FlashCopy licensing, I can mirror LUNs at the SAN level. Our link between sites is a 1000 mbps fiber connection.
My plan is to run one MTA server and one POA server at the primary site, with a POA server at the BC/DR site. Each would be attached to mirrored LUNs on the SANs. Backup servers would be in place for each POA and MTA and be set up in two-node clusters each, for 3 two-node clusters total. POA1 would fail to the BC/DR site, to the POA1 backup server attached to mirrored POA1 LUN, for example.
As long as our link is fast enough and our primary server room VLAN is extended to our BC/DR site, this should be feasible, correct?
I was also thinking about one WebAccess Agent and one WebAccess APP server at each site. That would be app and agent on the same server at each site. I'm not sure if I can set these up to load balance each other with DNS, or if I'd need a load balance appliance.
Does this sound like overkill for the amount of users I'm talking about? Is there a better way to plan for near real-time failover if an entire site is down?
And here's the response from Tim Heywood ...
GW is very IO intensive, so unless the sync between the SANs is synchronous (dark fibre) the second SAN will always be behind and cause issues. Having said that, having a second POA configured to run at the DR site - but not running unless the primary site is down - would probably work.
An alternative and cheaper solution would be to try using Reload to act as the DR solution. Reload is from GWAVA.
As for the WA side, I would implement that as 2 webaccess agents and 2 applications and cross-link them. Then if one goes down (agent or application), the other will pick up the load. In the event of a site loss, it would already be in place and working.
Novell Cool Solutions (corporate web communities) are produced by WebWise Solutions. www.webwiseone.com