Skip to content

Commit c34057c

Browse files
JamesJBarnettJamesJBarnett
authored
Merge pull request #247404 from s-polly/stp_known-issues_8-4-23
AzureML known issues
2 parents 9eca73d + 84d6876 commit c34057c

10 files changed

+357
-39
lines changed

‎articles/machine-learning/known-issues/application-sharing-policy-not-supported.md

Lines changed: 41 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,41 @@
1+
---
2+
title: Known issue - Application Sharing Policy isn't supported
3+
titleSuffix: Azure Machine Learning
4+
description: Configuring the applicationSharingPolicy property for a compute instance has no effect
5+
author: s-polly
6+
ms.author: scottpolly
7+
ms.topic: troubleshooting
8+
ms.service: machine-learning
9+
ms.subservice: core
10+
ms.date: 08/14/2023
11+
ms.custom: known-issue
12+
---
13+
14+
# Known issue - The ApplicationSharingPolicy property isn't supported for compute instances
15+
16+
Configuring the `applicationSharingPolicy` property for a compute instance has no effect as that property isn't supported
17+
18+
19+
[!INCLUDE [dev v2](../includes/machine-learning-dev-v2.md)]
20+
21+
**Status:** Open
22+
23+
**Problem area:** Compute
24+
25+
26+
## Symptoms
27+
28+
When creating a compute instance, the documentation lists an `applicationSharingPolicy` property with the options of:
29+
30+
- Personal only the creator can access applications on this compute instance.
31+
- Shared, any workspace user can access applications on this instance depending on their assigned role.
32+
33+
Neither of these configurations have any effect on the compute instance.
34+
35+
## Solutions and workarounds
36+
37+
There's no workaround as this property isn't supported. The documentation will be updated to remove reference to this property.
38+
39+
## Next steps
40+
41+
- [About known issues](azure-machine-learning-known-issues.md)

‎articles/machine-learning/known-issues/azure-machine-learning-known-issues.md

Lines changed: 36 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,36 @@
1+
---
2+
title: Azure Machine Learning known issues
3+
description: Identify issues that are affecting Azure Machine Learning features.
4+
author: s-polly
5+
ms.author: scottpolly
6+
ms.topic: troubleshooting
7+
ms.service: machine-learning
8+
ms.subservice: core
9+
ms.date: 08/04/2023
10+
ms.custom:
11+
---
12+
13+
# Azure Machine Learning known issues
14+
15+
This page lists known issues for Azure Machine Learning features. Before submitting a Support request, review this list to see if the issue that you're experiencing is already known and being addressed.
16+
17+
18+
## Currently active known issues
19+
20+
Select the **Title** to view more information about that specific known issue.
21+
22+
23+
|Area |Title |Issue publish date |
24+
|---------|---------|---------|
25+
|Compute | [Jupyter R Kernel doesn't start in new compute instance images](jupyter-r-kernel-not-starting.md) | August 14, 2023 |
26+
|Compute | [Provisioning error when creating a compute instance with A10 SKU](compute-a10-sku-not-supported.md) | August 14, 2023 |
27+
|Compute | [Idleshutdown property in Bicep template causes error](compute-idleshutdown-bicep.md) | August 14, 2023 |
28+
|Compute | [Slowness in compute instance terminal from a mounted path](compute-slowness-terminal-mounted-path.md)| August 14, 2023|
29+
|Compute| [Creating compute instance after a workspace move results in an Etag conflict error.](workspace-move-compute-instance-same-name.md)| August 14, 2023 |
30+
31+
32+
## Next steps
33+
34+
35+
- [See Azure service level outages](https://azure.status.microsoft/status)
36+
- [Get your questions answered by the Azure Machine Learning community](https://learn.microsoft.com/answers/tags/75/azure-machine-learning)

‎articles/machine-learning/known-issues/compute-a10-sku-not-supported.md

Lines changed: 32 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,32 @@
1+
---
2+
title: Known issue - A10 SKU not supported
3+
titleSuffix: Azure Machine Learning
4+
description: While trying to create a compute instance with A10 SKU, users encounter a provisioning error.
5+
author: s-polly
6+
ms.author: scottpolly
7+
ms.topic: troubleshooting
8+
ms.service: machine-learning
9+
ms.subservice: core
10+
ms.date: 08/14/2023
11+
ms.custom: known-issue
12+
---
13+
14+
# Known issue - Provisioning error when creating a compute instance with A10 SKU
15+
16+
While trying to create a compute instance with A10 SKU, you'll encounter a provisioning error.
17+
18+
:::image type="content" source="media/compute-a10-sku-not-supported/ci-a10.png" alt-text="A screenshot showing the provisioning error message.":::
19+
20+
[!INCLUDE [dev v2](../includes/machine-learning-dev-v2.md)]
21+
22+
**Status:** Open
23+
24+
**Problem area:** Compute Instance
25+
26+
## Solutions and workarounds
27+
28+
A10 AKUs aren't supported for compute instances. Consult this list of supported SKUs: [Supported VM series and sizes](https://learn.microsoft.com/azure/machine-learning/concept-compute-target?view=azureml-api-2#supported-vm-series-and-sizes&preserve-view=true)
29+
30+
## Next steps
31+
32+
- [About known issues](azure-machine-learning-known-issues.md)

‎articles/machine-learning/known-issues/compute-idleshutdown-bicep.md

Lines changed: 41 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,41 @@
1+
---
2+
title: Known issue - Compute | Idleshutdown property in Bicep template causes error
3+
titleSuffix: Azure Machine Learning
4+
description: When creating an Azure Machine Learning compute instance through Bicep compiled using MSBuild NuGet, using the `idleTimeBeforeShutdown` property as described in the API reference results in an error.
5+
author: s-polly
6+
ms.author: scottpolly
7+
ms.topic: troubleshooting
8+
ms.service: machine-learning
9+
ms.subservice: core
10+
ms.date: 08/04/2023
11+
ms.custom: known-issue
12+
---
13+
14+
# Known issue - Idleshutdown property in Bicep template causes error
15+
16+
When creating an Azure Machine Learning compute instance through Bicep compiled using [MSBuild/NuGet](../../azure-resource-manager/bicep/msbuild-bicep-file.md), using the `idleTimeBeforeShutdown` property as described in the API reference [Microsoft.MachineLearningServices workspaces/computes API reference](/azure/templates/microsoft.machinelearningservices/workspaces/computes?pivots=deployment-language-bicep) results in an error.
17+
18+
19+
20+
[!INCLUDE [dev v2](../includes/machine-learning-dev-v2.md)]
21+
22+
23+
**Status:** Open
24+
25+
26+
**Problem area:** Compute
27+
28+
## Symptoms
29+
30+
When creating an Azure Machine Learning compute instance through Bicep compiled using [msbuild/nuget](../../azure-resource-manager/bicep/msbuild-bicep-file.md), using the `idleTimeBeforeShutdown` property as described in the API reference [Microsoft.MachineLearningServices workspaces/computes API reference](/azure/templates/microsoft.machinelearningservices/workspaces/computes?pivots=deployment-language-bicep) results in an error.
31+
32+
33+
## Solutions and workarounds
34+
35+
To allow the property to be set, you can suppress warnings with the `#disable-next-line` directive. Enter `#disable-next-line BCP037` in the template above the line with the warning:
36+
37+
:::image type="content" source="media/compute-idleshutdown-bicep/disable-next-line.png" alt-text="Screenshot depicting the use of the #disable-next-line directive.":::
38+
39+
## Next steps
40+
41+
- [About known issues](azure-machine-learning-known-issues.md)

‎articles/machine-learning/known-issues/compute-slowness-terminal-mounted-path.md

Lines changed: 52 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,52 @@
1+
---
2+
title: Known issue - Slowness in compute instance terminal from a mounted path
3+
titleSuffix: Azure Machine Learning
4+
description: While using the compute instance terminal inside a mounted path of data folder, any commands executed from the terminal result in slowness.
5+
author: s-polly
6+
ms.author: scottpolly
7+
ms.topic: troubleshooting
8+
ms.service: machine-learning
9+
ms.subservice: core
10+
ms.date: 08/04/2023
11+
ms.custom: known-issue
12+
---
13+
14+
# Known issue - Slowness in compute instance terminal from a mounted path
15+
16+
While using the compute instance terminal inside a mounted path of a data folder, any commands executed from the terminal result in slowness. This issue is restricted to the terminal; running the commands from SDK using a notebook works as expected.
17+
18+
19+
[!INCLUDE [dev v2](../includes/machine-learning-dev-v2.md)]
20+
<!--- Choose the correct include --->
21+
22+
**Status:** Open
23+
24+
**Problem area:** Compute
25+
26+
## Symptoms
27+
28+
While using the compute instance terminal inside a mounted path of a data folder, any commands executed from the terminal result in slowness. This issue is restricted to the terminal; running the commands from SDK using a notebook works as expected.
29+
30+
### Cause
31+
32+
The `LD_LIBRARY_PATH` contains an empty string by default, which is treated as the current directory. This causes many library lookups on remote storage, resulting in slowness.
33+
34+
As an example:
35+
36+
```python
37+
LD_LIBRARY_PATH /opt/intel/compilers_and_libraries_2018.3.222/linux/mpi/intel64/lib:/opt/intel/compilers_and_libraries_2018.3.222/linux/mpi/mic/lib::/anaconda/envs/azureml_py38/lib/:/usr/local/cuda/lib64:/usr/local/cuda/extras/CUPTI/lib64/
38+
```
39+
40+
Notice the `::` in the path. This is the empty string, which is treated as the current directory.
41+
42+
When one of the paths in a list is "" - every executable tries to find the dynamic libraries it needs relative to current working directory.
43+
44+
## Solutions and workarounds
45+
46+
On the CI set the path making sure that `LD_LIBRARY_PATH` doesn't contain an empty string.
47+
48+
```export LD_LIBRARY_PATH="$(echo $LD_LIBRARY_PATH | sed 's/\(:\)\1\+/\1/g')"```
49+
50+
## Next steps
51+
52+
- [About known issues](azure-machine-learning-known-issues.md)

‎articles/machine-learning/known-issues/jupyter-r-kernel-not-starting.md

Lines changed: 57 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,57 @@
1+
---
2+
title: Known issue - Compute instance | Jupyter R Kernel doesn't start in new compute instance images
3+
titleSuffix: Azure Machine Learning
4+
description: When trying to launch an R kernel in JupyterLab or a notebook in a new compute instance, the kernel fails to start
5+
author: s-polly
6+
ms.author: scottpolly
7+
ms.topic: troubleshooting
8+
ms.service: machine-learning
9+
ms.subservice: core
10+
ms.date: 08/14/2023
11+
ms.custom: known-issue
12+
---
13+
14+
# Known issue - Jupyter R Kernel doesn't start in new compute instance images
15+
16+
When trying to launch an R kernel in JupyterLab or a notebook in a new compute instance, the kernel fails to start with `Error: .onLoad failed in loadNamespace()`
17+
18+
[!INCLUDE [dev v2](../includes/machine-learning-dev-v2.md)]
19+
20+
21+
**Status:** Open
22+
23+
24+
**Problem area:** Compute
25+
26+
27+
## Symptoms
28+
29+
After creating a new compute instance, try to launch R kernel in JupyterLab or a Jupyter notebook. The kernel fails to launch. You'll see the following messages in the Jupyter logs:
30+
31+
32+
```
33+
Aug 01 14:18:48 august-compute2Q6DP2A jupyter[11568]: Error: .onLoad failed in loadNamespace() for 'pbdZMQ', details:
34+
Aug 01 14:18:48 august-compute2Q6DP2A jupyter[11568]: call: dyn.load(file, DLLpath = DLLpath, ...)
35+
Aug 01 14:18:48 august-compute2Q6DP2A jupyter[11568]: error: unable to load shared object '/usr/local/lib/R/site-library/pbdZMQ/libs/pbdZMQ.so':
36+
Aug 01 14:18:48 august-compute2Q6DP2A jupyter[11568]: libzmq.so.5: cannot open shared object file: No such file or directory
37+
Aug 01 14:18:48 august-compute2Q6DP2A jupyter[11568]: Execution halted
38+
```
39+
40+
## Solutions and workarounds
41+
42+
To work around this issue, run this code in the compute instance terminal:
43+
44+
```azurecli
45+
jupyter kernelspec list
46+
47+
sudo rm -r <path/to/kernel/directory>
48+
49+
conda create -n r -y -c conda-forge r-irkernel jupyter_client
50+
conda run -n r bash -c 'Rscript <(echo "IRkernel::installspec()")'
51+
jupyter kernelspec list
52+
53+
```
54+
55+
## Next steps
56+
57+
- [About known issues](azure-machine-learning-known-issues.md)

‎articles/machine-learning/known-issues/media/compute-a10-sku-not-supported/ci-a10.png

45.6 KB
Loading

‎articles/machine-learning/known-issues/media/compute-idleshutdown-bicep/disable-next-line.png

143 KB
Loading

‎articles/machine-learning/known-issues/workspace-move-compute-instance-same-name.md

Lines changed: 39 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,39 @@
1+
---
2+
title: Known issue - After a workspace move, creating a compute instance with the same name as a previous compute instance will fail
3+
titleSuffix: Azure Machine Learning
4+
description: After moving a workspace to a different subscription or resource group, creating a compute instance with the same name as a previous compute instance will fail with an Etag conflict error.
5+
author: s-polly
6+
ms.author: scottpolly
7+
ms.topic: troubleshooting
8+
ms.service: machine-learning
9+
ms.subservice: core
10+
ms.date: 08/14/2023
11+
ms.custom: known-issue
12+
---
13+
14+
# Known issue - Creating compute instance after a workspace move results in an Etag conflict error.
15+
16+
After a moving a workspace to a different subscription or resource group, creating a compute instance with the same name as a previous compute instance will fail with an Etag conflict error.
17+
18+
19+
[!INCLUDE [dev v2](../includes/machine-learning-dev-v2.md)]
20+
<!--- Choose the correct include --->
21+
22+
**Status:** Open
23+
24+
**Problem area:** Compute
25+
26+
## Symptoms
27+
28+
After a workspace move, creating a compute instance with the same name as a previous compute instance will fail due to an Etag conflict error.
29+
30+
When you make a workspace move the compute resources aren't moved to the target subscription. However, you can't use the same compute instance names that you were using previously.
31+
32+
33+
## Solutions and workarounds
34+
35+
To resolve this issue, use a different name for the compute instance.
36+
37+
## Next steps
38+
39+
- [About known issues](azure-machine-learning-known-issues.md)

0 commit comments

Comments
 (0)