Provisioning from VMWare template deployed only as thick provisioned

Support
,
1

Problem: Hosts deployed from Image based provisioning on VMWare are deployed as thick provisioned even though “Thin Provision” is selected.

I am currently trying to create new hosts based on a VMWare template, everything works ok so far (thanks to reading through a lot of topics in this community). The only issue I am having is that the host is created with a “Thick Provisioned Lazy Zeroed” disk, meanwhile I am selecting “Thin Provision” on the Virtual Machine tab under storage configuration.

Expected outcome:
New VM should be deployed with a thin provisioned disk.

Foreman and Proxy versions: 1.16.1

Foreman and Proxy plugin versions:
foreman_setup 5.0.0

Other relevant data:
Currently using VMWare vCenter, with a vSAN cluster. Hypervisors version: VMware ESXi, 6.5.0, 7388607

Please let me know what log files should I paste that might help.

Regards.

2

We are also experiencing this exact issue, when we mark the image deployment as “thin provisioned” we still get a thick provisioned VM.

Versions
Vmware VCenter Server 6.0.0
Foreman: 1.16.1

3

Hi,

Can I get a foreman-debug from your foreman-instance so I can look further. We do have an open BZ on that downstream but I swear that was fixed not too long ago. Maybe it was and there is regression now, either way email me a foreman-debug at chrobert@redhat.com

4

Hello,

I sent you the foreman-debug via email.

I should also note, that the same issue is present on a freshly installed 1.17 version, in this case the patches specified here: Bug #23378: Unable to create a VM on 1.17.0 on vCenter 6.5. - Foreman where applied.

Thank you very much for your support!

5

@ekonomi Hi,

Thank you for the email, sorry I was away the last few days travelling. I will start to work on this tomorrow for you.

6

Thank you!

7

Are you using VSAN too or just a normal datastore cluster?

8

I am using VSAN. I checked the storage policy and it is set to thin-provision. When I try to deploy the VM from the same template on vCenter the machine i get is thin-provisioned as expected.

9

Thank you for clarifying, @lukayeh are you using VSAN too?

10

Just for some additional clarification. I used the same foreman instance to provision on an other ESXI host which has a local datastore and got the same results.

11

Ok good to know it is not isolated to VSAN, I will try with localstorage. I am just getting an image spun up and created. I do not have a VSAN license so thank you for letting me know that.

12

@ekonomi

I tested this on foreman 1.15 with fog-vsphere 1.7.0 and it worked fine:

http://nimb.ws/7bPd2f

I am going to test with nightly and see if we have regression.

13

Thank you for your time cintrix84.

Let me know if you need any additional information from me.

14

Hi @ekonomi

I went ahead and built a host normally:

http://nimb.ws/grU8qG

I made sure to have the vm created with thick provisioning

http://nimb.ws/ZjXvNs

Sadly this does not look to be working, as you can see from here the vm is still in thin mode.

http://nimb.ws/LXzZlI

It looks like we are hitting this:

https://bugzilla.redhat.com/show_bug.cgi?id=1378579

Did it ever work for you?

I can pick up this bug and start to look at it if you wish?

I tested a thick prov template and trying to make a new thin vm with that and it didnt work either:

Template: http://nimb.ws/L6rIW3

Thin checked: http://nimb.ws/MsBumD

Thin result: http://nimb.ws/xNEbR4

I tested a thin template and was able to create a thick vm ok

Thin thick create: http://nimb.ws/FtvS7Z

Thick result: http://nimb.ws/EBNzbt

When I tried to create a thin vm from a thin template it did not honor it:

Thin create: http://nimb.ws/SibMkE

Thin result: http://nimb.ws/0P3wO1

15

Thank you for running the tests @cintrix84

I remember trying the PXE Mode installation and the image was deployed as Thin-Provision. I didnt try setting it as thick provisioned though.

Im not sure if Bug:1378579 is directly related to our issue, since it looks like it has the opposite outcome where the image is deployed as thin provisioned.

However I would really appreciate if you can pick up this bug.

Thank you very much @cintrix84, I really appreciate it!

16

Another data point here, I can confirm this problem also exists in 1.18. I am unable to deploy a VM template using image as thin when the template is thick.

17

Sorry for the long delay, I am finishing up 1 high priority bug for our downstream release and then will start to look at this more. I already have it assigned to me to look at.

18

Looks like I am impacted by this as well. Technically I’m running Satellite, but it worked correctly before a recent upgrade. I am going back through my recently provisioned hosts to try and figure out exactly when it stopped working correctly.

19

I upgraded to 1.20.1 and this problem went away. However, I am reasonably certain this was not related to the upgrade (could be wrong). My solution involved tweaking the POST for the VM creation to match the example someone provided in Bug #25093: Provisioning VMs with Foreman API Not Working for Multiple SCSI Controllers and Volumes - Hammer CLI - Foreman

The VM attributes are very picky about data entry. The exact before and after are below… and in my opinion… what I had before should have worked and it didn’t. Synopsis below,

What I had before,

                            "compute_attributes":

                            {

                                            "scsi_controllers":

                                            [

{ “type”:“VirtualLsiLogicController”, “key”:“1000” }
],

                                            "volumes_attributes":

                                            {

                                                            "0":

{ “thin”:1, “eagerZero”:0, “mode”:“persistent”, “controllerKey”:1000, “datastore”:“XXXXXX” }
}

                                            "start":"1",

                                            "image_id":"5034a01f-5f5c-7411-44f7-c3c3c66264ca",

                                            "cpus":"2","corespersocket":"1",

                                            "memory_mb":"4096",

                                            "firmware":"bios",

                                            "resource_pool":"foreman-temp",

                                            "hardware_version":"Default",

                                            "add_cdrom":"1",

                            }

What I changed it to,

“compute_attributes”: {

"start": "1",

"image_id": "5034a01f-5f5c-7411-44f7-c3c3c66264ca",

"cpus": "8",

"corespersocket": "1",

"memory_mb": "8192",

"firmware": "bios",

"resource_pool": "foreman-temp",

"hardware_version": "Default",

"add_cdrom": "0",

"scsi_controller_type": "ParaVirtualSCSIController",

"volumes_attributes": {

  "0":

{ “_delete”: “”, “datastore”: “XXXXXX”, “name”: “Hard disk 1”, “size_gb”: “150”, “thin”: “true”, “eagerZero”: “false” }
}

},

20

I upgraded to 1.20.1 today and nothing changed for me, a new VM from a thin provision image was created with lazy zeroed storage.
Can you explain your old workaround?
Thanks