Azure-docs: Support for Nvidia GPU Images
Can I deploy https://azuremarketplace.microsoft.com/en-us/marketplace/apps/nvidia.ngc_azure_17_11?tab=PlansAndPrice using scale sets using low priority?
Document Details
⚠Do not edit this section. It is required for docs.microsoft.com ➟ GitHub issue linking.
- ID: 7c353c07-72b3-9a2b-2e4b-a79d6b60f07a
- Version Independent ID: 90b738c8-fe64-1657-d834-432c91d93a5b
- Content: Azure virtual machine scale sets overview
- Content Source: articles/virtual-machine-scale-sets/overview.md
- Service: virtual-machine-scale-sets
- GitHub Login: @gatneil
- Microsoft Alias: negat
carlfm01
All 14 comments
Getting plan error using the portal, also for https://azuremarketplace.microsoft.com/en-gb/marketplace/apps/microsoft-ads.linux-data-science-vm-ubuntu?tab=PlansAndPrice
carlfm01
on 11 Oct 2018
Thanks for the feedback! We are currently investigating and will update you shortly.
Karishma-Tiwari-MSFT
on 11 Oct 2018
@carlfm01 For your second question, I tried creating through the portal and it is successfully deployed. Need you input on the following:
- Can you please confirm that you tried creating it through the portal?
- please share the subscription plan you have?
@gatneil Need your help in answering customer's first question if that is supported. Can you please take a look?
Karishma-Tiwari-MSFT
on 11 Oct 2018
Hello @Karishma-Tiwari-MSFT, sorry for the delay, if you tag me in a edit the Github email notification system doesn't send the notification of the new tag.
Yes, using the portal.
My suscription: Microsoft BizSpark without expending limit, offer ID: MS-AZR-0064P.
Vm type: NC6 v2 single GPU
Region: South Central US
For the firts type of Linux image I'm experiencing the half of performance with exact dependencies compared to Windows. That's why I want to deploy the other types of vm before I debug the vm causing the performance issue.
I can deploy :
I can't deploy:
I can't deploy:
carlfm01
on 11 Oct 2018
Here's the deploy.json that the portal generates, I removed the suscription ids.
{
"id": "/providers/Microsoft.Resources/deployments/microsoft.vmss-20181011144201",
"name": "microsoft.vmss-20181011144201",
"properties": {
"templateHash": "16003712594483244245",
"parameters": {
"location": {
"type": "String",
"value": "southcentralus"
},
"vmssName": {
"type": "String",
"value": "lvmpricing"
},
"vmSku": {
"type": "String",
"value": "Standard_NC6s_v2"
},
"adminUsername": {
"type": "String",
"value": "neoxz"
},
"instanceCount": {
"type": "String",
"value": "1"
},
"image": {
"type": "Object",
"value": {
"sku": "linuxdsvm",
"publisher": "microsoft-ads",
"version": "1.7.3",
"offer": "linux-data-science-vm"
}
},
"singlePlacementGroup": {
"type": "String",
"value": "true"
},
"pipName": {
"type": "String",
"value": ""
},
"pipLabel": {
"type": "String",
"value": ""
},
"skuType": {
"type": "String",
"value": "Standard"
},
"ipAllocationMethod": {
"type": "String",
"value": "Static"
},
"priority": {
"type": "String",
"value": "Low"
},
"subnetId": {
"type": "String",
"value": "/providers/Microsoft.Network/virtualNetworks/vn/subnets/default"
},
"enableAcceleratedNetworking": {
"type": "String",
"value": "false"
},
"publicIpAddressPerInstance": {
"type": "String",
"value": "true"
},
"upgradeMode": {
"type": "String",
"value": "Manual"
},
"evictionPolicy": {
"type": "String",
"value": "Deallocate"
},
"adminPassword": {
"type": "SecureString"
},
"vnetName": {
"type": "String",
"value": "vn"
}
},
"mode": "Incremental",
"debugSetting": {
"detailLevel": "RequestContent, ResponseContent"
},
"provisioningState": "Running",
"timestamp": "2018-10-11T21:02:26.4047832Z",
"duration": "PT7.8573342S",
"correlationId": "a7f7a546-15c7-4e20-b207-8c50a54b7ec6",
"providers": [
{
"namespace": "Microsoft.Network",
"resourceTypes": [
{
"resourceType": "virtualNetworks",
"locations": [
"southcentralus"
]
},
{
"resourceType": "networkSecurityGroups",
"locations": [
"southcentralus"
]
}
]
},
{
"namespace": "Microsoft.Compute",
"resourceTypes": [
{
"resourceType": "virtualMachineScaleSets",
"locations": [
"southcentralus"
]
}
]
}
],
"dependencies": [
{
"dependsOn": [
{
"id": "/providers/Microsoft.Network/networkSecurityGroups/lvmpricingnsg",
"resourceType": "Microsoft.Network/networkSecurityGroups",
"resourceName": "lvmpricingnsg"
}
],
"id": "/providers/Microsoft.Network/virtualNetworks/vn",
"resourceType": "Microsoft.Network/virtualNetworks",
"resourceName": "vn"
},
{
"dependsOn": [
{
"id": "/providers/Microsoft.Network/virtualNetworks/vn",
"resourceType": "Microsoft.Network/virtualNetworks",
"resourceName": "vn"
}
],
"id": "/providers/Microsoft.Compute/virtualMachineScaleSets/lvmpricing",
"resourceType": "Microsoft.Compute/virtualMachineScaleSets",
"resourceName": "lvmpricing"
}
],
"validationLevel": "Template"
}
}
Also failing for the dsvm using CentOS, hope the json helps.
carlfm01
on 11 Oct 2018
@carlfm01 yes, you should be able to deploy that VM image to a low-pri scale set.
The error message about "plan" means that you need to:
1) Accept the pricing for that VM image in the portal
2) provide the "plan" object in your ARM template (or other deployment method)
It's a tad involved, so I wrote up a blog post on how to do this: https://negatblog.wordpress.com/2018/10/13/azure-os-images-what-is-a-plan/
That being said, it sounds like you're deploying from the portal, which means the portal should be including the "plan" object in the request. @Karishma-Tiwari-MSFT this probably requires a fix from the person/team that publishes the NVIDIA image. I have no idea who publishes that image, but @BorisB2015 might :)
@mmccrory as FYI on low-pri scale set usage
Cheers! :)
-Neil
gatneil
on 13 Oct 2018
(via email)
@ankitsud - you have debugged this error message for me before (see comments in the GitHub thread - missing plan parameter when provisioned via portal):
see: https://github.com/MicrosoftDocs/azure-docs/issues/16638#issuecomment-428853578
This refers to the following image being deployed via portal:
"image": { "type": "Object", "value": { "sku": "linuxdsvm", "publisher": "microsoft-ads", "version": "1.7.3", "offer": "linux-data-science-vm" } }
Could you please comment if this is a known bug (I believe I have not seen it on the images I work with/publish and where I have seen this message before - I wonder if it is a regression of any kind).
Mine was a "loose VM", not a VMSS, and I wonder if this is the cause of the problem.
Thanks,
Boris.
BorisB2015
on 14 Oct 2018
Thanks @gatneil, I enabled both images in the portal but still getting the same error. I tried using the portal but failed, then edited the deploy to add the plan and imagereference still failing.
For nvidia image I used, still the same error.
"imageReference": {
"publisher": "nvidia",
"offer": "ngc_azure_17_11",
"sku": "nvidia_gpu_cloud_18_06",
"version": "18.09.1"
}
"plan": {
"name": "nvidia_gpu_cloud_18_06",
"publisher": "nvidia",
"product": "ngc_azure_17_11"
},
I will try with Ubuntu image soon.
Cheers,
-Carlos.
carlfm01
on 14 Oct 2018
@ankitsud Umm, there was a bug in the portal which the team fixed some time ago, this should never be hit while deploying an image from the portal.
Can you confirm that?
From: Boris Baryshnikov
Sent: Saturday, October 13, 2018 4:01 PM
To: MicrosoftDocs/azure-docs azure-docs@noreply.github.com; MicrosoftDocs/azure-docs reply@reply.github.com; Ankit Sud Ankit.Sud@microsoft.com
Cc: Mention mention@noreply.github.com
Subject: Re: [MicrosoftDocs/azure-docs] Support for Nvidia GPU Images (#16638)
(via email)
@Ankit SudAnkit.Sud@microsoft.com - you have debugged this error message for me before (see comments in the GitHub thread - missing plan parameter when provisioned via portal):
[cid:[email protected]]
This refers to the following image being deployed via portal:
"image": { "type": "Object", "value": { "sku": "linuxdsvm", "publisher": "microsoft-ads", "version": "1.7.3", "offer": "linux-data-science-vm" } }
Could you please comment if this is a known bug (I believe I have not seen it on the images I work with/publish and where I have seen this message before - I wonder if it is a regression of any kind).
Mine was a "loose VM", not a VMSS, and I wonder if this is the cause of the problem.
Thanks,
Boris.
From: Neil Sant Gat <[email protected]notifications@github.com>
Sent: Saturday, October 13, 2018 1:31:39 AM
To: MicrosoftDocs/azure-docs
Cc: Boris Baryshnikov; Mention
Subject: Re: [MicrosoftDocs/azure-docs] Support for Nvidia GPU Images (#16638)
@carlfm01https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fcarlfm01&data=02%7C01%7CAnkit.Sud%40microsoft.com%7C77f2a1afa8e24586883308d6315fb5d6%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C636750684473462827&sdata=Uxl%2B%2BPuJstNJl0rLRCJBt%2FLmbFwUpQY30TcdqfJ9RK8%3D&reserved=0 yes, you should be able to deploy that VM image to a low-pri scale set.
The error message about "plan" means that you need to:
- Accept the pricing for that VM image in the portal
- provide the "plan" object in your ARM template (or other deployment method)
It's a tad involved, so I wrote up a blog post on how to do this: https://negatblog.wordpress.com/2018/10/13/azure-os-images-what-is-a-plan/https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fnegatblog.wordpress.com%2F2018%2F10%2F13%2Fazure-os-images-what-is-a-plan%2F&data=02%7C01%7CAnkit.Sud%40microsoft.com%7C77f2a1afa8e24586883308d6315fb5d6%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C636750684473472835&sdata=PbsE6h0xMgP1PVLjA4dh4UL3Etpyk%2FWMptRGQB%2FAbKQ%3D&reserved=0
That being said, it sounds like you're deploying from the portal, which means the portal should be including the "plan" object in the request. @Karishma-Tiwari-MSFThttps://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FKarishma-Tiwari-MSFT&data=02%7C01%7CAnkit.Sud%40microsoft.com%7C77f2a1afa8e24586883308d6315fb5d6%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C636750684473482844&sdata=ty2zevIq5uWPpTzvr2iIhKuNYooOG9zCpaWzh263%2BUA%3D&reserved=0 this probably requires a fix from the person/team that publishes the NVIDIA image. I have no idea who publishes that image, but @BorisB2015https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FBorisB2015&data=02%7C01%7CAnkit.Sud%40microsoft.com%7C77f2a1afa8e24586883308d6315fb5d6%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C636750684473492852&sdata=boMNFZgXsEv6Jcur3NHha6VrS4r0OeKnFWs%2BwF6qTKE%3D&reserved=0 might :)
@mmccroryhttps://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fmmccrory&data=02%7C01%7CAnkit.Sud%40microsoft.com%7C77f2a1afa8e24586883308d6315fb5d6%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C636750684473492852&sdata=olXidmovBuLG3OoDCSAmivaDFK3PGATktwAjOMsn9Jk%3D&reserved=0 as FYI on low-pri scale set usage
Cheers! :)
-Neil
-
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHubhttps://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FMicrosoftDocs%2Fazure-docs%2Fissues%2F16638%23issuecomment-429522447&data=02%7C01%7CAnkit.Sud%40microsoft.com%7C77f2a1afa8e24586883308d6315fb5d6%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C636750684473502856&sdata=lxkgCOWWk6vL6tluGpxvP5f%2BvPcR8%2FCo34ZKUu2jB%2Fw%3D&reserved=0, or mute the threadhttps://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAK1G0DTUTB6DLVeQWS70eqyL6Nz96JXjks5ukaTrgaJpZM4XW89S&data=02%7C01%7CAnkit.Sud%40microsoft.com%7C77f2a1afa8e24586883308d6315fb5d6%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C636750684473512868&sdata=Tt8CHQ6bLVFUUxD3fyXaLPdk1JY1rC%2Bqs2rh6agFhdI%3D&reserved=0.
BorisB2015
on 15 Oct 2018
@carlfm01 Thanks for bringing this to our attention. I am working offline with the product team on this issue and will share updates here.
Karishma-Tiwari-MSFT
on 19 Oct 2018
@carlfm01 The product team is working on the fix. The PR will be merged soon. Thanks for bringing this to our attention. We will close this issue now. I will share here once the changes are live.
Karishma-Tiwari-MSFT
on 23 Oct 2018
Hi @Karishma-Tiwari-MSFT, any news?
-Carlos.
carlfm01
on 30 Oct 2018
hi @carlfm01 I checked the status of the PR for the fix of this issue. It will go out in the next week's build as shared by the product team. Once that build is out, it would be out in all public regions the following week. I will keep you updated. Thanks for the patience.
Karishma-Tiwari-MSFT
on 30 Oct 2018
@carlfm01 The fix has been deployed to production in all regions. You can test it out now. Thanks for your patience.
Karishma-Tiwari-MSFT
on 28 Nov 2018
Related issues
bdcoder2
·
3Comments
jebeld17
·
3Comments
jharbieh
·
3Comments
mrdfuse
·
3Comments
Agazoth
·
3Comments
Most helpful comment
@carlfm01 yes, you should be able to deploy that VM image to a low-pri scale set.
The error message about "plan" means that you need to:
1) Accept the pricing for that VM image in the portal
2) provide the "plan" object in your ARM template (or other deployment method)
It's a tad involved, so I wrote up a blog post on how to do this: https://negatblog.wordpress.com/2018/10/13/azure-os-images-what-is-a-plan/
That being said, it sounds like you're deploying from the portal, which means the portal should be including the "plan" object in the request. @Karishma-Tiwari-MSFT this probably requires a fix from the person/team that publishes the NVIDIA image. I have no idea who publishes that image, but @BorisB2015 might :)
@mmccrory as FYI on low-pri scale set usage
Cheers! :)
-Neil