CGiuliani
Contributor
Contributor

Is there any way to prevent SRM from starting all VMs in a recovery plan at once - Priority groups don't work if you use IP Customization

I have about 50 VMs configured in SRM 5.8, and I'd like the have one recovery plan that can be used to start everything up at as quickly as possible.  But I need to have some delays between the VMs to spread out the startup workload.  I thought I could use priority groups for this.  There can be only five priority groups, but I thought that would be spread out the workload enough to avoid timeouts in service startups or in waiting for VMware tools, etc.  The problem is that IP customization, which includes a startup and shutdown of the guest VM, is not affected by priority groups.  In the case of a recovery plan test, every VM startup includes seven sub-steps:  1) Configure Storage 2) Configure Test network 3) Guest Startup 4) IP customization 5) Guest Shutdown 6) Power On 7) Wait for VMware Tools.  Priority groups only affect the last two sub-steps.

So, not only is the description of this feature very deceptive about what it does, it gives me no way to avoid my hosts and storage from getting slammed with 50 simultaneous VM startups (and shutdowns.)

VMware support suggested I use the base VM boot delay to introduce delays for each VM.   That could work, but would also affect ever restart in my production environment.  Does anybody have any better ideas?  Does anybody else agree this feature should be modified to include the IP customization steps?

Tags (1)
0 Kudos
1 Reply
CGiuliani
Contributor
Contributor

After I pushed them a bit more, VMware support pointed me to  KB 2013167, which has exactly what I needed.  There are some SRM advanced settings which can only be set by editing the VMware-dr.xml file:  defaultMaxBootAndShutdownOpsPerCluster and defaultMaxBootAndShutdownOpsPerHost.  I used defaultMaxBootAndShutdownOpsPerCluster and it did exactly what I needed.

0 Kudos