Staggered vApp Startup question

Deathmage

Hey guys,

So the VCP exam is on Wednesday of next week and I'm playing around with the lab and I've got a general best practice kind of question.

So I have a vApp which lets just say has an allocation of 8000 MHz CPU and 12 GB's of RAM inside of that is two Resource pools (each pool has 3000Mhz limit with 1300MHz per server reservation and a 3800 GB limit with a 1600 GB reservation), one called "Existence" which has a DC (2008 R2) and a member server with SQL 2008 R2 and the other resource pool called "Honeypot" that has the FTP and Exchange Server (Both Ubuntu Server 14.04-based but technically speaking irrelevant). Obviously I set the Existence pool to fully come online with a 3 minute holding pattern before the Honeypot even initializes since the DC has DNS on it and the SQL database both which are dependent upon for FTP/Exchange to work properly.

My question is this, I set the DC to boot 1st with a 120 sec pause before SQL even turns on, is that too short or should 2 minutes be enough time for the DC to post, and load all of it's services like AD, DNS, Group Policy. Even during posting the CPU caps out at 900 mhz maybe then drops down to like 100 mhz hence why the reservation is only 1300mhz, why waste the resources?

From power on to Ctrl+Alt+del on the lab is less than 1:30 seconds; the iSCSI switch has a throughput of 196mbps and MTU set to 9412 on all ends of the fabric so the data transfer is blazing fast; the NAS has 5 WD Red drives in a RAID 6. I'm sure this question is all dependent upon the iOPS of the storage fabric in a production environment but I'm curious if anyone has a rough number from there experience with resource pools being used in this manner inside a vApp?

Essendon

In a lab 2 minutes should be enough I reckon for the DC to go from from off to the logon screen. Increase the pause for SQL to 180 seconds if you are concerned. Why have reservations at all? Let the machines grab what they need and let the hypervisor decide on how to dish the honey out. I have never used resource pools in vApps at all, in fact I am an opponent of resource pools because of the management overhead they introduce.

For your exam next week, take a look at the resource pool questions in our 'your daily VMware quiz' thread. Expect to be asked about pools in the exam! And good luck!

Deathmage

Essendon wrote: »

In a lab 2 minutes should be enough I reckon for the DC to go from from off to the logon screen. Increase the pause for SQL to 180 seconds if you are concerned. Why have reservations at all? Let the machines grab what they need and let the hypervisor decide on how to dish the honey out. I have never used resource pools in vApps at all, in fact I am an opponent of resource pools because of the management overhead they introduce.

For your exam next week, take a look at the resource pool questions in our 'your daily VMware quiz' thread. Expect to be asked about pools in the exam! And good luck!

Thanks!

I keep scoring 75% on the VMware practise exams, would be nice if those explanation told you why it was wrong and what was right. Been scowing over the very long guides from the blueprint for the past few weeks. I've realized at this point it's so hard to remember everything so I'm just doing my best to read them all and get as much of a grasp on things as I can.....also 2nd guessing myself seems to get the questions wrong, dam 'inner-voice'!

See I can build-out a Cluster with up to 4 hosts with vCenter integration perfectly fine, example the cluster I just made on Saturday, it's gone all week and no problems, in-fact customer is calling me all the time asking questions it's working so well, lol! - my biggest issue is the case scenario questions that I honestly just don't have the in-depth experience with.... figuring out the questions is fun, but I'm figuring I got 30 seconds per question on the exam so it' a fun 30 seconds, sigh!

dave330i

Essendon wrote: »

I have never used resource pools in vApps at all, in fact I am an opponent of resource pools because of the management overhead they introduce.

RPs have their uses. I like to use them to collapse clusters. If you make the RP multiples of your ESXi host resources, the management becomes easy.

Essendon

I'll agree with this use of resource pools. Gotta keep in mind in though is by default RPs can only use upto 94% of the cluster's resources. I'd just leave everything at the root and let the hosts decide what they want to do.

dave330i

Essendon wrote: »

I'll agree with this use of resource pools. Gotta keep in mind in though is by default RPs can only use upto 94% of the cluster's resources. I'd just leave everything at the root and let the hosts decide what they want to do.

I don't remember reading the 94%. Do you have a source you can link? Thx.

Edit: Found couple of blogs about it, but couldn't find a VMware doc about it.

Deathmage

gosh I was fiddling around with the vApp some more and I wanted to break it but I'm now trying to figure out what I did to break it, to the point that I had to reboot the host because shutdown didn't do **** for the VM's manually... here's some screenies, can someone enlighten me?

Deathmage

here is a map:

Deathmage

hmmm... I changed one value from 0 seconds to 10 seconds and now the vApp is starting in the delayed series I wanted to too.... but now to figure out why!!!

All I changed: 0 secs to 10 secs...

SimonD.

I would suggest it's starting up because the previous VM is reporting VMware Tools is ready and then it's moving on to the next machine (the same kind of behaviour occurs in SRM as well).

Deathmage

SimonD. wrote: »

I would suggest it's starting up because the previous VM is reporting VMware Tools is ready and then it's moving on to the next machine (the same kind of behaviour occurs in SRM as well).

ahhh that makes sense... missed that...