container instance no response sometimes

Question

container instance no response sometimes

Chris Yang 20

Hi,

We deployed a web app (Mageai) on the container instance, but sometimes the web just hangs up loading and results in no response. And at the same time if you open the portal, you can see the cpu and network drop to 0. And maybe after a few hours, the web will be back online. We also check the container events, during the 'no response' time, there were no 'killing' and 'started' events.

That happens very frequently; it may work today, but fail to load tomorrow. Could you give us some suggestions on how to troubleshoot it?

Chris Yang 20 Reputation points

2025-11-10T23:46:40.6433333+00:00

and it backs online again, you can see from the dashboard, the server showed no response and then back online after couple of minutes.
Himanshu Shekhar 1,935 Reputation points Microsoft External Staff Moderator

2025-11-11T16:01:40.8033333+00:00
Hello Chris Yang - Thank you for reaching Microsoft Qna Platform

During the web app hang event, have you captured detailed container logs or Mage AI application logs? Are there any error messages or warnings recorded inside the container logs at the time the CPU and network drop to zero?

Has the container instance ever experienced Out-of-Memory (OOM) events or resource throttling? What is the current CPU and memory allocation for the container and do memory or CPU metrics show any spikes coinciding with that hang period?

Are there any known external service dependencies (databases, APIs, storage accounts) from Mage AI that could be causing the application to wait indefinitely?

Have you checked network connectivity, DNS resolution, and endpoint availability during the hang occurrences?

Is there a health monitoring probe configured for the Mage AI container in Azure? If not, please consider adding application-level health checks and configuring restart policies to automatically recover from hung states?
Chris Yang 20 Reputation points

2025-11-13T00:44:18.7+00:00
Hi Himanshu,

Thank you so much for the quick response, and I am sorry for the late reply. Here are the answers.

During the web app hang event, have you captured detailed container logs or Mage AI application logs? Are there any error messages or warnings recorded inside the container logs at the time the CPU and network drop to zero?

Yes, I have checked the log. There was no error on the logs either from the app or the container. The log stayed on whenever the app worked ok. There was no new log when the issue came out.

Has the container instance ever experienced Out-of-Memory (OOM) events or resource throttling? What is the current CPU and memory allocation for the container and do memory or CPU metrics show any spikes coinciding with that hang period?
`No out-of-memory occurred. its a 16gb memory container, and during the hang period, the memory stayed very constant 2.25gb. Only CPU and network drop to zero.`

Are there any known external service dependencies (databases, APIs, storage accounts) from Mage AI that could be causing the application to wait indefinitely?

the container mounts volumes on a blob storage, they are in the same resource group.

Have you checked network connectivity, DNS resolution, and endpoint availability during the hang occurrences?

no

Is there a health monitoring probe configured for the Mage AI container in Azure? If not, please consider adding application-level health checks and configuring restart policies to automatically recover from hung states?

maybe not. could you share me an instruction of how to set it up?
Himanshu Shekhar 1,935 Reputation points Microsoft External Staff Moderator

2025-11-14T12:54:17.57+00:00

@Chris Yang

The liveness probe detects when the application is in an unhealthy state and triggers automatic restarts https://learn.microsoft.com/en-us/azure/container-instances/container-instances-liveness-probe#yaml-deployment

If MageAI uses a different framework, adapt the endpoint accordingly. Also refer for Health probes in Azure Container Apps : https://learn.microsoft.com/en-us/azure/container-apps/health-probes?tabs=arm-template
Chris Yang 20 Reputation points

2025-11-14T18:11:48.84+00:00

Hi Himanshu,

Thanks for the reply. I just realized that problem is similar to the cold start. But, I dont find a place to make it always on.
Himanshu Shekhar 1,935 Reputation points Microsoft External Staff Moderator

2025-11-17T15:49:41.9566667+00:00

@Chris Yang

Unlike App Services, ACI containers can be deallocated on idle which cause cold starts on the next request when the container has to be pulled, started, and initialized again.

This process delays or hangs if the container or app initialization takes long. There is no built-in "Always On" toggle in ACI to prevent this.

If you are keeping the container always warm is critical then Azure Container Apps, which supports "Always On" through minimum replicas configuration, reducing cold starts.

Reducing cold-start time on Azure Container Apps https://learn.microsoft.com/en-us/azure/container-apps/cold-start

Since, MageAI uses mounted blob storage, network or storage access issues can cause the application to freeze while waiting for data.

During these hang periods, monitor network connectivity, DNS resolution, and the health of the storage account. - https://learn.microsoft.com/en-us/azure/container-instances/container-instances-troubleshooting

1 answer

Your answer

Chris Yang 20 Reputation points

2025-11-10T23:46:40.6433333+00:00

and it backs online again, you can see from the dashboard, the server showed no response and then back online after couple of minutes.
Himanshu Shekhar 1,935 Reputation points Microsoft External Staff Moderator

2025-11-11T16:01:40.8033333+00:00

Hello Chris Yang - Thank you for reaching Microsoft Qna Platform

During the web app hang event, have you captured detailed container logs or Mage AI application logs? Are there any error messages or warnings recorded inside the container logs at the time the CPU and network drop to zero?

Has the container instance ever experienced Out-of-Memory (OOM) events or resource throttling? What is the current CPU and memory allocation for the container and do memory or CPU metrics show any spikes coinciding with that hang period?

Are there any known external service dependencies (databases, APIs, storage accounts) from Mage AI that could be causing the application to wait indefinitely?

Have you checked network connectivity, DNS resolution, and endpoint availability during the hang occurrences?

Is there a health monitoring probe configured for the Mage AI container in Azure? If not, please consider adding application-level health checks and configuring restart policies to automatically recover from hung states?
Chris Yang 20 Reputation points

2025-11-13T00:44:18.7+00:00

Hi Himanshu,

Thank you so much for the quick response, and I am sorry for the late reply. Here are the answers.

During the web app hang event, have you captured detailed container logs or Mage AI application logs? Are there any error messages or warnings recorded inside the container logs at the time the CPU and network drop to zero?

Yes, I have checked the log. There was no error on the logs either from the app or the container. The log stayed on whenever the app worked ok. There was no new log when the issue came out.

Has the container instance ever experienced Out-of-Memory (OOM) events or resource throttling? What is the current CPU and memory allocation for the container and do memory or CPU metrics show any spikes coinciding with that hang period?
`No out-of-memory occurred. its a 16gb memory container, and during the hang period, the memory stayed very constant 2.25gb. Only CPU and network drop to zero.`

Are there any known external service dependencies (databases, APIs, storage accounts) from Mage AI that could be causing the application to wait indefinitely?

the container mounts volumes on a blob storage, they are in the same resource group.

Have you checked network connectivity, DNS resolution, and endpoint availability during the hang occurrences?

no

Is there a health monitoring probe configured for the Mage AI container in Azure? If not, please consider adding application-level health checks and configuring restart policies to automatically recover from hung states?

maybe not. could you share me an instruction of how to set it up?
Himanshu Shekhar 1,935 Reputation points Microsoft External Staff Moderator

2025-11-14T12:54:17.57+00:00

@Chris Yang

The liveness probe detects when the application is in an unhealthy state and triggers automatic restarts https://learn.microsoft.com/en-us/azure/container-instances/container-instances-liveness-probe#yaml-deployment

If MageAI uses a different framework, adapt the endpoint accordingly. Also refer for Health probes in Azure Container Apps : https://learn.microsoft.com/en-us/azure/container-apps/health-probes?tabs=arm-template
Chris Yang 20 Reputation points

2025-11-14T18:11:48.84+00:00

Hi Himanshu,

Thanks for the reply. I just realized that problem is similar to the cold start. But, I dont find a place to make it always on.
Himanshu Shekhar 1,935 Reputation points Microsoft External Staff Moderator

2025-11-17T15:49:41.9566667+00:00

@Chris Yang

Unlike App Services, ACI containers can be deallocated on idle which cause cold starts on the next request when the container has to be pulled, started, and initialized again.

This process delays or hangs if the container or app initialization takes long. There is no built-in "Always On" toggle in ACI to prevent this.

If you are keeping the container always warm is critical then Azure Container Apps, which supports "Always On" through minimum replicas configuration, reducing cold starts.

Reducing cold-start time on Azure Container Apps https://learn.microsoft.com/en-us/azure/container-apps/cold-start

Since, MageAI uses mounted blob storage, network or storage access issues can cause the application to freeze while waiting for data.

During these hang periods, monitor network connectivity, DNS resolution, and the health of the storage account. - https://learn.microsoft.com/en-us/azure/container-instances/container-instances-troubleshooting

Answer 1

Chris Yang

Unlike App Services, ACI containers can be deallocated on idle which cause cold starts on the next request when the container has to be pulled, started, and initialized again.

This process delays or hangs if the container or app initialization takes long. There is no built-in "Always On" toggle in ACI to prevent this.

If you are keeping the container always warm is critical then Azure Container Apps, which supports "Always On" through minimum replicas configuration, reducing cold starts.

Reducing cold-start time on Azure Container Apps https://learn.microsoft.com/en-us/azure/container-apps/cold-start

Since, MageAI uses mounted blob storage, network or storage access issues can cause the application to freeze while waiting for data.

During these hang periods, monitor network connectivity, DNS resolution, and the health of the storage account. - https://learn.microsoft.com/en-us/azure/container-instances/container-instances-troubleshooting

Share via

container instance no response sometimes

1 answer

Your answer