Good afternoon all. I have a question/issue with Composer that I need some advice on before I call up support.
This is my first few times working with Composer and linked clones. It seems that once I deploy a resource pool, VCenter creates the replica and source machines right away, but hangs at 94% progress before deploying machines to the pool. It takes on average about two hours for it to clear and eventually deploy the pool.
I've read some things that say this could be related to DHCP not being released - to the VMs size being too big. Right now I'm trying to deploy a pool of VMs from an 80GB parent with 20GB user data partitions.
Should I call support to work with this issue or can anyone give me advice as to whether or not that's expected when deploy VMs of that size.
Thanks!
Andrew Orr
My eval gold image was a beefy (for us, at least) 16GB and our iSCSI LUN was set to an 8k default. Eval provisions with this config took about 20-25 min and also had a hang at 94%.
Once I got serious and slimmed down the image to < 4GB and set the LUN to 64k my provision time dropped to 3 min.
It's just a guess but I believe the 94% complete on the UI refers to 94% of tasks completed not 94% of time complete (completion percentage is not time weighted).
...I'd suggest provisioning a tiny test image and starting a timer. I bet you see a difference.
It's not a bug or issue. That 94% hang is the actual part of when the cloning of the VMDK from the Master image into the replica is done. Obviously, the larger the VMDK the longer that pause will be.
Now I do have one question: why 80GB master with 20GB of UDD? Why not use network shares for the user data and reduce the size of the Guest OS?
Want make a difference in the future of VMware Products? Feature request your ideas ( )!
Would increased vHardware specs to my VCenter server make a difference or is it basically expected to initialize the pool in this timeframe regardless.
The 80GB OS drive is for our developers to have multiple installs of Visual Studio, SQL Managment Studio, Powerbuilder...etc. The 20GB is to host their own private data aside from the network share. I've reduced that from 20GB to 2GB and seen no difference in the deployment of the pool.
It's not the UDD that hangs it, it's the 80GB copy. The reality is that your storage, vCenter and other factors are what's killing it. Given the size, it's going to present initial challenges for deployment. Given what they are doing it might have performance hits. What kind of storage are you using (make/model/etc.)
Want make a difference in the future of VMware Products? Feature request your ideas ( )!
It's an HP MSA 2Gbps fibre san, active active connection.
I've deployed a second pool across another HP EVA SAN with the same results. That seems to rule out our first SAN as the culprit. Any other thoughts?
For how long does it hang? I think it is more an ui issue. It will take some time to deploy a full clone depending on the size.
Sent with my BlackBerry
For info, with VMware View 3.1 and a WinXP Pro SP3 master of 10GB (Storage is a NetApp FAS3020 using iSCSI LUNs), the cloning from Master to Replica takes (94% phase) takes 7 minutes.
Is that fast for you or not?
@Christoph - the deploy process hangs at 94% for about 120 minutes with an 80GB master.
@Erik - thanks for the feedback on your experience. I'm trying to deploy with a 10GB master this afternoon to see how my time compares. I'll keep you posted.
I think the 94% issue is more just an issue of the UI. In a performing environment a 10 GB - 15 GB will take 10-15 minutes. You should try to clone your 80 GB image outside View just in vCenter and get the time for that.
I had this issue until I did the Following
On the Virtual Center Server - go to services of the Windows Server
Find the VMware Universal File Access Service - stop the service and make sure it is set to Manual..
I have only had the issue arise when the service is set to Automatic.
your replica that was stuck at 94% should complete quickly, and then the rest of that pool should finish faster.
The next pool you create should not run into the same issue.
Not sure the why the UFA service needs to be Manual over Automatic, but this has made a world of difference in our environment.
Hi Andrew,
Are you using pool persistent or noon persintent?
Remove disk d and use only disk c.
If you are using pool persistent, use the option redirection data users.
Regards
My eval gold image was a beefy (for us, at least) 16GB and our iSCSI LUN was set to an 8k default. Eval provisions with this config took about 20-25 min and also had a hang at 94%.
Once I got serious and slimmed down the image to < 4GB and set the LUN to 64k my provision time dropped to 3 min.
It's just a guess but I believe the 94% complete on the UI refers to 94% of tasks completed not 94% of time complete (completion percentage is not time weighted).
...I'd suggest provisioning a tiny test image and starting a timer. I bet you see a difference.
I agree with AUPhil, 94% of the task is complete its just waiting for the copy of the replica is complete.
It appears that this is a normal wait time, I'm doing a deployment with composer enabled and the initial replica of the parent that gets put onto each Datastore. The task quickly jumps up to 94%, only to hang there until the replica has been cloned onto the datastore. We're using an EMC Clariion CX4-120 on 10K Fibre Channel disks. It's a new deployment with only the one pool on it and we still have the same issue. It makes sense because the replica is a copy of the VM that gets stored on each datastore where your linked clones reside. Each datastore will contain linked clones that reference the replica on the datastore the same datastore that each linked clone is stored on.
-Brad