October 25, 2010 at 9:11 am
Hi there - can anyone advise on the potential issues / pitfalls with the following environment.
2 node Cluster (Active / Passive) both machines VM's running on seperate physical hosts. Shared SAN storage between 2 servers to allow for Microsoft clustering & SQL2005 to be installed and configured for failover. Data availability does not appear to be the primary factor here with Application availability being the most important thing.....but in my opinion the 2 go hand in hand. Its no use having an applicaton that you can logon if there's no data behind it.
I believe that using VMware and Clustering can cause issues with regards to patching, Data high availability, application availability etc but looking to get an overall picture of the pros and cons of both.
Yes VMware snapshot technology is a nice to have for the likes of patching etc - however my understanding this that this does not work with shared disk storage and Microsoft clustering. Is this the case?
Data availability - well snapshots of the data can be taken at a SAN level but is this the best method to employ.
Would a Single VM machine running SQL server and possibly deploying SQL Mirroring be a valid option here and give greater coverage incase of an outage. VM could be snapshot - data would be mirrored elsewhere - downtime and data loss kept at a min?
All opinions appreciated on this.
October 25, 2010 at 1:17 pm
I am curious to the replies to this topic as I am in the same boat myself. I do share the same opinion though that Data availability and Application availability goes hand in hand (by application, you mean the SQL Server services itself, not the applications connecting to it).
Yes VMware snapshot technology is a nice to have for the likes of patching etc - however my understanding this that this does not work with shared disk storage and Microsoft clustering. Is this the case?
I think I've read some articles that in order to do Microsoft clustering, the disks has to be presented as RDM (Physical compatibility). And if the disks are presented as RDM, the snapshot capability is disabled. but that's for the data. If the OS itself though is presented as VMDK, then patching, and snapshots can still be done, it just doesn't affect the disks that were RDM (p).
Data availability - well snapshots of the data can be taken at a SAN level but is this the best method to employ.
Would a Single VM machine running SQL server and possibly deploying SQL Mirroring be a valid option here and give greater coverage incase of an outage. VM could be snapshot - data would be mirrored elsewhere - downtime and data loss kept at a min?
These are the same idea that is playing on my thoughts: On having a single VM, I would recommend testing out the vMotion of this single VM to another host. But it seems that having a cluster is still the better idea, just in case the host go down, or the application going awry, or if you want to do some patching on that one VM, if you want to do some configuration changes, etc.
On another hand, deploying SQL mirroring along with clustering seems to cover both data availability and application availability.
I am still in the "Virtualize or Not" cross-road. Are you able to successfully implement SQL Server on a VM? (I would probably need to spin up another thread for this question)
October 26, 2010 at 6:55 am
Hi and thanks for taking the time to reply to my intial question.
RAW Disk Mappings are indeed a hinderance with VMware and Microsoft clustering. From what I have read I agree with your reply to this.
We've now gone away from the "Microsoft Clustering" scenario (with less then a month to go-live!!!) and will be using the VMware High Availability tools instead along with Database Mirroring. This allows for the VM's to be snapshot prior to service packs etc, make availability of the SQL application better as well although it means we need to look into possibly a tool for monitoring the Windows services etc as VMware wont recognise there is an issue with the services.
As far as VMware and SQL - I previously worked for a company where their whole SQL estate was virtualised (over 100 server) with multiple databases in excess of 400GB on each of the instances and had no issues what so ever. Backups of the databases were done using SQL LiteSpeed to disk and VM backups were taken as well so we could always get the data back. Its only when clustering comes into play it get complicated.
Although from what I've read Microsoft clustering will work but its not generally supported which kind of makes our decision for us as its no use having an unsupported environment.
hope this helps.
Viewing 3 posts - 1 through 2 (of 2 total)
You must be logged in to reply to this topic. Login to reply