vCenter 6.0.0 requires 8 GB's: no if and's or buts...

So this this evening I decided to update the home-cluster to 6.0.0 from 5.5 and gosh what an annoyance. you can no longer force vCenter onto a machine with less than 9 GB's of RAM, as it needs true 8196 to install and not all memory is true 8196.

So I'm over here P2V'in the vCenter to the home cluster, peg it to 12 GB's of RAM, power it on and then upgrading it to 6.0.0 and then throwing it back onto the host in a V2P. If vCenter doesn't like it screw it, it's just for a home lab.

I like having vCenter seperate from the cluster, call me nutso.

This might be helpful for others...

Find more posts tagged with

Comments

kj0

You can install and then scale back. However, I have noticed that some services may struggle to start if under 8Gb. (Although, could be part of a disk issue as well for my lab - Yet to be confirmed).

jibbajabba

You lost me at separate vCenter

Mike7

jibbajabba wrote: »

You lost me at separate vCenter

That is why @Deathmage is a VMware vExpert.

Deathmage

I don't like having all my cookies in one jar, I also have one physical DC in the home cluster in a RAID 1.

kj0 wrote: »

You can install and then scale back. However, I have noticed that some services may struggle to start if under 8Gb. (Although, could be part of a disk issue as well for my lab - Yet to be confirmed).

the Dell Optiplex, even though it had 8 GB's of RAM it appearred as 8046 so the installation wouldn't happen. It completed it's P2V around 2 am, not bad about an hour and 45 minutes. Now to power it on the ESXi host and upgrade it.

Just make a note, even if you have 8 GB's of RAM in a physical host if it's not 'true' 8196 it won't install.

Here is the error:

vCenter insufficent memory error.jpg

kj0

Yeap.... If you don't have vCenter in a Cluster, than what happens if that host goes down? What happens to your Availability of vCenter?

Deathmage

kj0 wrote: »

Yeap.... If you don't have vCenter in a Cluster, than what happens if that host goes down? What happens to your Availability of vCenter?

For a home-lab it works better for me. if the host goes down I have a spare mobo, and PSU for it plus all my home lab equipment is connected to a pair of APC UPS SMC1000-2U's. I've worked in the dark during and outage for 45 minutes with the cluster powered on and then the power came back on. I have 1 hour and 35 minutes of run time with my cluster and if I have my laptop attached to the front-facing power strip.

Also I have an Acronis image of the host, would take me ~40 minutes tops to reimage the PC in the event of a failure, I got plenty of spare hard drives and memory in the closet. My closet looks like a small grocery store but for nerds.

Basically to answer your question, I plan ahead.

However in a production world I'd run vCenter inside of a cluster since that cluster is 'always on' my home-cluster isn't 'always on', plus Veeam is on this station for all of my home-lab backups.

TheProf

I feel the same way... I have an HP Microserver N40L with 16GB of memory and windows server 2012 installed on it. I use it as my vCenter server. At some point though, vCenter is going to be virtualized in my lab, just that I need to get another host to build my second vCenter cluster (management cluster).

Deathmage

holly molly, vCenter 6.0.0 is a resource hog. I've had to bump this sucker to 12 GB's of ram and 6 vcpu's for this 2008 R2 VM to startup smoothly, but I'm stilll wating a solid 20 minutes for it ot load into memory. However the loading time might be the RAID 5 array on the NAS and the painfully slow nature of SQL Express loading into memory.

Pupil

kj0 wrote: »

You can install and then scale back. However, I have noticed that some services may struggle to start if under 8Gb. (Although, could be part of a disk issue as well for my lab - Yet to be confirmed).

I thought I was the only one experiencing this bug. Often I have to manually start some services because I'm using less than 8GB

jibbajabba

Mike7 wrote: »

That is why @Deathmage is a VMware vExpert.

Me, too

Seriously though, I get it. In fact, i used to have a micro server as vcenter which was a jump box at the same time. Just messin'

There are use cases for everything

Deathmage

well that P2V went well but then after upgrading the server on the ESXi host I spend a good 5 hours today literally just trying to log into vCenter on the ESXi host through the web client or the normal client and no luck, and the CPU was taxed at 100% for 6 vcores for a good 2+ hours.

I just blew away the VM and doing a fresh VM of 2008 R2 from a template, and just going to re-install vCenter 6 from scratch.

Not sure why the VM one tanked from the P2V, my only theory is cause it used to be on the hard drive it was on a normal installation and then when it was P2V'd it went to a Thick Lazy Zero and not a Eager Zero and since SQL Express was on it, it really tanked in performance.

So the new VM mentioned above I waited the 40 minutes for it to make a Thick Eager Zero Partition on the new VM in hopes this will solve that issue since I bumped it to 12 GB's of RAM and 6 vcpu's and it made no difference to it's 2+ hour and going 'loading' time...as someone else mentioned above I even went in safe mode and disabled every non-essential service and it made no change.

I honestly can't recall that happening when I upgraded from 5.1 to 5.5 so that's a new one for me...

Update: it might have been the Lazy Zero, a fresh install of vCenter 6 Standard went without a hitch, doing VUM atm and getting used to the Web Client. Really liked the other client, feels more fluid, the web client is sluggish.

jibbajabba

"6 vCPU" ... That may have been counter productive as well. Would have been interesting to peak at esx top to see the READY and CO STOP values of the VM. Unless your physical host as a ridiculous amount of course, you may ended up with a lot of scheduling.

jibbajabba

* ridiculous amount of cores

Deathmage

jibbajabba wrote: »

"6 vCPU" ... That may have been counter productive as well. Would have been interesting to peak at esx top to see the READY and CO STOP values of the VM. Unless your physical host as a ridiculous amount of course, you may ended up with a lot of scheduling.

The servers normally only start with 2 vcpu's but the performance was hammered at 100% on 2 or 6 cpus so it didn't help.

...was at 6 to see if it would help, was a troubleshooting step only.

6 core xeons with HT, two sockets per host.

Found an alternative use for the old vCenter server, using the Optiplex as a secondary DC and RDC point from my Gaming PC so I don't need to join the domain and run at root-level access, plus since it has a integrated GPU on the i7 on the Optiplex I can run my resolution to 1920 x 1080 in RDC for a much larger vSphere Web Client.

jibbajabba wrote: »

Me, too

Seriously though, I get it. In fact, i used to have a micro server as vcenter which was a jump box at the same time. Just messin'

There are use cases for everything

I like thinking outside of the box.

J.Tot

Boots fine for me at 5GB but slow
I give it 6GB and I have no problems, decent speed too.

TheProf

In the process of updating my lab with Update 1.. vCenter already done and luckily flawless upgrade. Well done VMware, well done! For once I didn't have to worry about services not starting properly post update.

Deathmage

it was a interesting update not having to worry about something failing on the upgrade, getting used to the web client takes some patience though.