May 25, 2017 at 11:13 am
Dear Colleagues,
How to demostrate, that one solution based on vm high availability is not the only one solution for a previoulsy clustered database (microsoft fail over instances), I have received a virtual machine from my vm administrators as to test SQL Server 2014 in a lab is compatible with his recommendation and clustered are no needed no more, so the point for me is that we are losing availability, I'm interested in make some tests that he can recognize my point of view . I know that there are lots of recommendations and best practices from Microsoft and from Vmware, but this guy is a little bit obfuscated.
1.) I already have explained that we will lose availability,
2.) SQL Server does not work with snapshots this is a panacea for my vm admin.
3.) Etc.
Any ideas that I could use to argue or to test in a lab ? I have hear that SQL Server can corrupt databases under some circuntamces with vmare, how I could demostrate this.
I have done a vmware cluster lab and now I'm working with a Always On lab, but I'm not a very good salesman
Thanks again for your ideas
May 25, 2017 at 11:26 am
luismarinaray - Thursday, May 25, 2017 11:13 AMDear Colleagues,How to demostrate, that one solution based on vm high availability is not the only one solution for a previoulsy clustered database (microsoft fail over instances), I have received a virtual machine from my vm administrators as to test SQL Server 2014 in a lab is compatible with his recommendation and clustered are no needed no more, so the point for me is that we are losing availability, I'm interested in make some tests that he can recognize my point of view . I know that there are lots of recommendations and best practices from Microsoft and from Vmware, but this guy is a little bit obfuscated.
1.) I already have explained that we will lose availability,
2.) SQL Server does not work with snapshots this is a panacea for my vm admin.
3.) Etc.
Any ideas that I could use to argue or to test in a lab ? I have hear that SQL Server can corrupt databases under some circuntamces with vmare, how I could demostrate this.I have done a vmware cluster lab and now I'm working with a Always On lab, but I'm not a very good salesman
Thanks again for your ideas
Instead of fighting against it, embrace the new possibilities it has IN your favor.
Maybe ask for 2 x VMs, set up an Availability Group, and argue 1 x Server does not provide HA on DB or OS Level, but AG does.
Moving to VMs, you have a lot more to gain than lose.
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
This thing is addressing problems that dont exist. Its solution-ism at its worst. We are dumbing down machines that are inherently superior. - Gilfoyle
May 26, 2017 at 8:52 am
Easy to test. Have a process, SQLCMD batch works fine, that constantly adds data, maybe every second. Crash VM, see where you are with data. Or see how long to restart/rebuild and see where the snapshot is.
I assume you have data on other disks outside the VM, and you could have log backups, but restore those. Show the time to restore. HA is about reducing RTO, so show that.
The talk RPO, data loss. How much data lost in snapshots or with backups. There is likely some, which clustering can help with assuming your disk isn't the failure. It could be, so clusters don't necessarily help here.
Figure out for sure what you are trying to protect against and the impact. Make sure you understand, then talk to the VMWare people. As Henrico said, a set of VMs is also good. I embrace virtualization, but understand what you get and lose.
Viewing 3 posts - 1 through 2 (of 2 total)
You must be logged in to reply to this topic. Login to reply