DPM and DRS

[gunnarb]

I've been using DPM and DRS for a few months now and just love watching my servers power on and off as resources are needed. I was trying something new the other day and I was wondering if anyone else has seen this. I created a rule to seperate two servers lets call them DC1 and DC2. I power on DC1 and it goes to the single ESX that is running (there is only one host running becuase only one host is needed presently). WHen I power DC2 on the second ESX has to come out of standby, once it is out of standby DC2 powers up... in a word, perfect.

When you power down DC2 all VMs will evacuate the host (DRS put the VMs there to keep it load balanced). Oh that reminds me I ignore HA requirements to allow DPM to power off all but one ESX host, I know this means in a failure it will take longer for the servers to post but I've tested this and the difference is 3 minutes (with two hosts) verse 10 minutes with one host. 10 minutes meets SLA so I'd like to conserver power as I'm a treehugger. In any case, all my VMs evacuate the host and go onto one ESX host. This would be fine if I wanted the server down for any extended period of time but I've see it do this on a reboot. I've seen it where 3 hosts are running and regardless it will boot a fourth ESX host just to but DC2 on it. It seems no matter what, after a shutdown of DC2 the VM will not power back up unless it manually boots its own server, as if the rule stated to boot on a host with no VMs running.

Has anyone experienced something like this? I know I'm pushing DRS and DPM here but my results have been great, this is my first weird issue, and it's not show stopping its just weird.

Gunnar

[wpatton]

Gunnarb,

Do you have any reservations set within the configuration of DC2's resources tab? If so, it might be too high and require additional resources are spun up to cover the Reservation, i.e. the second ESX host.

Also, do you have any Rules configured under the cluster DRS configuration that specify keeping Virtual Machines apart?

If you found this or other information useful, please consider awarding points for "Correct" or "Helpful".

[bulletprooffool]

HJave you got anything like processor affinity set on VMs? Or rules trying to keep them on certain Hosts?

also, verify that your HA requirements don;t have a specific host set as a recovery host, and that it allows for you to run without the extra redunndancy.

[AntonVZhbankov]

Since anti-affinity rule prevent DC2 from running on ESX1 with DC1, DRS powers ESX2 to satisfy requirement.

That's all.

---

VMware vExpert '2009

http://blog.vadmin.ru

[gunnarb]

I agree, my rule is designed to keep them on seperate boxes, however that doesn't make sense if two boxes are already powered up. THis is an exact scenario I had:

2 Hosts running with more than enough power to handle DC2 on either box.

DC2 is not running

I power up DC2

a 3rd ESX host powers up.

DC2 waits for the third host to power up then powers on on this third host.

Now here is the fun part, after DC2 is powered up, DRS migrates all the VMs off of the 2nd host and powers it down.

This is what I find weird. Its not a problem, it's just weird and I was wondering if anyone else has experienced this.

[AntonVZhbankov]

Create Service Request, I think there is some problem in DPM / DRS code. It should not work this way.

---

VMware vExpert '2009

http://blog.vadmin.ru

[gunnarb]

It's just an annoyance right now, so I'm skip opening a Service Request. I'm sure they'll get it figured out, I know DPM just exited experimental stages so they probably have some bugs left to fix.

Thanks for your feedback!

Gunnar

[AntonVZhbankov]

If you consider any comment as helpful, please award points 🙂
---

VMware vExpert '2009

http://blog.vadmin.ru

[gunnarb]

I don't think I can, I didn't put this up as a question, I posted as a discussion. Sorry man, I'd hook you up if I could.

[AntonVZhbankov]

C'est la vie!

---

VMware vExpert '2009

http://blog.vadmin.ru