how to achieve zero downtime with VMware for critical application

[prashantntm]

We do have case where the customer is looking for zero-downtime solution

for their critical applications. The solution we approached with is,

VMware HA & FT for the critical application. But these two

technologies only offers the high availability of virtual machines &

they are unaware of issue related to application services. In that case

we looked for application level clustering as well. Here we would like

to know how much application level clustering is supported along with

VMware HA & FT.

I went through a white paper which talks about symantec ApllicationHA to

be used for such environment. Does this solution guarantee the zero

transaction loss for critical application.

[NuggetGTR]

dont think anyone can guarantee zero downtime, for every 9 you add the cost exponentially grows tho.

Application HA sounds good, I dont have any experience with it but all that does is restart a service(s) so if there was an issue and worst case restarts the server, you would still get some downtime. not sure how this would work in a clustered solution or if it even does, but either way its not going to be zero downtime

FT and MSCS dont work together. and FT comes with the limitation of only allowing 1vcpu so that limits the larger apps i guess.

MSCS with VMware HA is a pretty good combo and same with NLB clusters if they are hardware load balanced. but again there is always the failover point when your offline weather its application HA restarting the service to restore it, or MSCS failing over resource, of HA powering back up a machine etc etc etc.

[idle-jam]

We use either NeverFail and Maraton everRun solutions for zero down time for our customers. Perhaps you could look into it.

iDLE-jAM | VCP 2, VCP 3 & VCP 4

If you found this or any other answer useful please consider the use of the Helpful or correct buttons to award points

[js411]

Symantec's ApplicationHA only monitors the health state of the app, and then restarts the VM if it fails. That will achieve HA, but not zero downtime.

Neither will a FT solution like Marathon or VMware FT, they run a single OS in lock-step across 2 machines... which provides zero downtime for just a very narrow set of uncommon scenarios... like if your server has a hard complete failure. But if you have a software failure, they do nothing.

As stated above, it will likely take a combination of technologies to achieve the best solution. And the less downtime a solution delivers, the higher the price tag will be. So while naive managers like to state lofty goals of zero downtime, when they see the price tag... they usually think twice. Beyond the intial cost, you need to ensure you have the IT staff to maintain a really complex system... there are many variables. So you need to establish the SLA's up front.

[depping]

If you absolutely need to guarantee 100% uptime I think you shouldn't be looking for a workaround which MSCS / HA / VM Monitoring / App Monitoring essentially is. All these solutions provide a mechanism to increase resiliency which wasn't added by the developer. The only way to guarantee no data loss would be a combination of of the mentioned workaround but even more important a well architected application. Distributed Databases combined with multiple App and/or Web servers would definitely add another 9. I suggest to first start looking into frameworks like Springsource.

Duncan

VMware Communities User Moderator | VCDX

-

---

Soon to be release: vSphere 4.1 HA and DRS deepdive (end of November through Amazon.com)

Blogging: | Twitter:

[prashantntm]

Hi champs,

Thanks for all your post.

Will keep you posted if i get futher development on the same.

Thanks,

Prashant

[AndrewBarnes]

Disclosure: I work for Neverfail

Some other things to consider when looking at Zero Downtime would include

1) Impact of storage failure: VMWare HA/FT, Clustering, Symantec ApplicationHA all assume that the storage is shared and will not fail. This needs to be considered.

2) Planned maintenance

3) Site failure, not just a "disaster" but also consider power outages etc

4) Physical systems, are all applications running in VMs or do existing physical servers need to be considered?

Maybe consider that VMware's vCenter Server Heartbeat availability module for vCenter Server is built on Neverfail technology and this gives an indication of how to address downtime.

Regards

Andrew