You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@airflow.apache.org by GitBox <gi...@apache.org> on 2022/07/28 14:54:35 UTC
[GitHub] [airflow] karkyra opened a new issue, #25372: CPU utilization 100 % loads in one worker nodes
karkyra opened a new issue, #25372:
URL: https://github.com/apache/airflow/issues/25372
### Apache Airflow version
2.1.1
### What happened
There are 7 workers. 1 worker node loads 100% CPU utilization, 5 worker nodes CPU Utilization 70-80 % and 1 worker nodes CPU utilization within 20 %. it doesn't distributing loads equally between the nodes.
Airflow running in Kubernetes and using airflow Kubernetes executor.
### What you think should happen instead
Loads should be distributed equally between worker nodes
### How to reproduce
N/A
### Operating System
redhat 8.2
### Versions of Apache Airflow Providers
N/A
### Deployment
Docker-Compose
### Deployment details
_No response_
### Anything else
_No response_
### Are you willing to submit PR?
- [ ] Yes I am willing to submit a PR!
### Code of Conduct
- [X] I agree to follow this project's [Code of Conduct](https://github.com/apache/airflow/blob/main/CODE_OF_CONDUCT.md)
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@airflow.apache.org.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [airflow] plutoid commented on issue #25372: CPU utilization 100 % loads in one worker nodes
Posted by GitBox <gi...@apache.org>.
plutoid commented on issue #25372:
URL: https://github.com/apache/airflow/issues/25372#issuecomment-1217269462
hi,
@potiuk , at last I use py-spy instead to capture the processes for flame graph , kindly have a look, thanks a lot!
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@airflow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [airflow] plutoid commented on issue #25372: CPU utilization 100 % loads in one worker nodes
Posted by GitBox <gi...@apache.org>.
plutoid commented on issue #25372:
URL: https://github.com/apache/airflow/issues/25372#issuecomment-1214178060
I'm still struggling with issue that pyflame not support py3.7, since airflow now running on py3.7.
https://github.com/uber-archive/pyflame/issues/120
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@airflow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [airflow] potiuk commented on issue #25372: CPU utilization 100 % loads in one worker nodes
Posted by GitBox <gi...@apache.org>.
potiuk commented on issue #25372:
URL: https://github.com/apache/airflow/issues/25372#issuecomment-1228931382
Sorry for delay - but it Looks like vast majority of time is exec ssh command - You seem to have an SSH hook that likely simply gets a lot of output to process and it it does so.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@airflow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [airflow] potiuk closed issue #25372: CPU utilization 100 % loads in one worker nodes
Posted by GitBox <gi...@apache.org>.
potiuk closed issue #25372: CPU utilization 100 % loads in one worker nodes
URL: https://github.com/apache/airflow/issues/25372
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@airflow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [airflow] plutoid commented on issue #25372: CPU utilization 100 % loads in one worker nodes
Posted by GitBox <gi...@apache.org>.
plutoid commented on issue #25372:
URL: https://github.com/apache/airflow/issues/25372#issuecomment-1250316382
thanks for your reply, I will check with my colleagues and see what will happen if we direct output to /dev/null.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@airflow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [airflow] potiuk commented on issue #25372: CPU utilization 100 % loads in one worker nodes
Posted by GitBox <gi...@apache.org>.
potiuk commented on issue #25372:
URL: https://github.com/apache/airflow/issues/25372#issuecomment-1210555788
> hi,
>
> I'm using airflow 2.3.2 in docker not Kubernetes , the worker nodes are consuming almost CPUs too, as per attachment, not sure what's doing!
>
> VMs: 4core 16GB nodes: 2
>
> any ideas for this? Many thanks! <img alt="810" width="555" src="https://user-images.githubusercontent.com/1091981/183844947-557cd362-4976-4940-af73-2bc655996e5d.png">
I thin the question is what kind of tasks your worker are running - but only you can tell it. Possibly you spin up some processes that simply consume a lot of CPU - you need to investigate further to see what your python processes are doing. This is not a widespread problem so likely this is the code you run as part of your taskt that are doing it.
You can likely use 'trace' built in module or I actually recommend pyflame https://pyflame.readthedocs.io/en/latest/usage.html to produce interactive flamechart. See the instructions on how to run it - but it can give answer what actually takes the time.
I woudl be very curious to see results of the investigation you do and flamechart.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@airflow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [airflow] potiuk commented on issue #25372: CPU utilization 100 % loads in one worker nodes
Posted by GitBox <gi...@apache.org>.
potiuk commented on issue #25372:
URL: https://github.com/apache/airflow/issues/25372#issuecomment-1198315718
Airflow is not a solution to distribute the load evenly. I am not sure where you took the notion that this "should" happen.
You can control your Kubernetes resources on your own if you want and fine-tune it, but Airflow does not do it (and will likely never do) on it's own.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@airflow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [airflow] plutoid commented on issue #25372: CPU utilization 100 % loads in one worker nodes
Posted by GitBox <gi...@apache.org>.
plutoid commented on issue #25372:
URL: https://github.com/apache/airflow/issues/25372#issuecomment-1210294470
hi,
I'm using airflow 2.3.2 in docker not Kubernetes , the worker nodes are consuming almost CPUs too, as per attachment, not sure what's doing!
VMs: 4core 16GB
nodes: 2
any ideas for this? Many thanks!
<img width="555" alt="810" src="https://user-images.githubusercontent.com/1091981/183844947-557cd362-4976-4940-af73-2bc655996e5d.png">
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@airflow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org