|
RSS Feed - WebSphere MQ Support
|
RSS Feed - Message Broker Support
|
 |
|
Business service restoration / fault tolerance / multi-inst |
« View previous topic :: View next topic » |
Author |
Message
|
lancelotlinc |
Posted: Thu Mar 03, 2011 5:48 am Post subject: Business service restoration / fault tolerance / multi-inst |
|
|
 Jedi Knight
Joined: 22 Mar 2010 Posts: 4941 Location: Bloomington, IL USA
|
Started this new thread to discuss the merits of different configurations. Some people like the Active-passive concept. Others prefer the Active-Active.
Since I have a preference for the latter, could someone make the case for the former?
What are the advantages of Active-Passive? _________________ http://leanpub.com/IIB_Tips_and_Tricks
Save $20: Coupon Code: MQSERIES_READER |
|
Back to top |
|
 |
bsiggers |
Posted: Thu Mar 03, 2011 9:23 am Post subject: |
|
|
Acolyte
Joined: 09 Dec 2010 Posts: 53 Location: Vancouver, BC
|
Depends on your goals, I guess. The main advantages of active-active is that you don't really have routine downtime, so if I'm say, for example, applying the latest security patches, I can remove one node, patch it up, then bring it back into the cluster, and then do the same with the other node - no downtime.
Whereas in an active-passive system, you are basically going to have some kind of downtime in a real-life scenario, whether you measure that in milliseconds, seconds or minutes.
For us, the main advantage of active-active is the lack of routine downtime. Better resource utilization is a side-benefit - but at the same time, you have to make sure you have enough spare capacity to tolerate a total loss of one of the nodes - so my question would be, are you really saving resources in this case?
However, DR is another issue - for that I think active-passive makes sense. I don't want to have traffic flowing over the WAN between two seperate datacentres if I can avoid it, and avoid paying for multiple licenses at the same time. But, of course, it depends on your requirements - I'm biased by my own specific needs. |
|
Back to top |
|
 |
lancelotlinc |
Posted: Thu Mar 03, 2011 9:29 am Post subject: |
|
|
 Jedi Knight
Joined: 22 Mar 2010 Posts: 4941 Location: Bloomington, IL USA
|
Thanks for the reply. Do you have an active-passive setup at this time? If so, when was the last time the active-passive was failed over to the alternate node? _________________ http://leanpub.com/IIB_Tips_and_Tricks
Save $20: Coupon Code: MQSERIES_READER |
|
Back to top |
|
 |
bsiggers |
Posted: Thu Mar 03, 2011 9:52 am Post subject: |
|
|
Acolyte
Joined: 09 Dec 2010 Posts: 53 Location: Vancouver, BC
|
We have an active-passive setup for DR, which is accomplished by having everything running on VMWare ESX, and using SAN replication to the DR site. So it doesn't use the MQ or Broker-specific facilities for setting up things active-passive. This is the DR strategy used for the whole enterprise, not just the middleware layer.
Have tested failover once, and plan to do it again in the future to make sure it can be done in the required SLA. Note this is relatively new for us. |
|
Back to top |
|
 |
lancelotlinc |
Posted: Thu Mar 03, 2011 10:08 am Post subject: |
|
|
 Jedi Knight
Joined: 22 Mar 2010 Posts: 4941 Location: Bloomington, IL USA
|
This is an illustrative point I would like to draw:
DR plans need to be tested regularly. I would advocate every day, not once a week, month or year.
The short coming I have noticed with Active-Passive setups is that more than half the time they fail to operate as intended when the time comes. Human intervention is many times required in order for the Passive node to come up. This delays the business, sometimes by hours.
By contrast, if you run Active-Active, you always know that your alternate node will work as designed in the event of an outage of the primary node, since it is working anyway.
This concept I find to be true no matter if we are talking about Message Broker or WAS or some other production system. Passive nodes never get the exercise they need, and if they are ever tested, the tests are not realistic scenarios or the yardstick by which success is measured is compromised to call it a successful test when, in reality, it failed the true test. _________________ http://leanpub.com/IIB_Tips_and_Tricks
Save $20: Coupon Code: MQSERIES_READER |
|
Back to top |
|
 |
bsiggers |
Posted: Thu Mar 03, 2011 10:49 am Post subject: |
|
|
Acolyte
Joined: 09 Dec 2010 Posts: 53 Location: Vancouver, BC
|
Hmmm... I guess it depends on how many disasters you get per year. Maybe your site is constantly surrounded by backhoe drivers looking for fiberoptic cables to cut, or angry mobs storm the datacenter daily - I don't judge
For example's sake - say our DR strategy is designed around a once-per-decade occurrence - then I'd say you should be testing your DR strategy annually, for example.
However, take a traditional scenario where you're taking outages based on some kind of monthly cycle (windows update, kernel patching, etc.) - how many outages do you have to take before you realize that an active-active is a better solution?
I'm a fundamentally lazy person, so my preference is always active-active wherever feasible in a normal production environment, but I see DR as something very seperate and different, where active-passive (and, of course, some human intervention) is acceptable for a once-per-decade type event. But of course, this will vary depending on your requirements. |
|
Back to top |
|
 |
|
|
 |
|
Page 1 of 1 |
|
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
|
|
|
|