GitHub user fengchutadie added a comment to the discussion: When start a dag in 
airflow 3.0.0, scheduler could be crashed but works well in airflow 2.10.4

it seems it's not about out of resources. 
in console log, part of the scheduler log as below
```
scheduler  | [2025-05-13T01:58:37.100+0000] {_client.py:1026} INFO - HTTP 
Request: PATCH 
http://localhost:8080/execution/task-instances/0196c1ed-1bfe-7616-961a-af093a67ca90/run
 "HTTP/1.1 504 Gateway Timeout"
scheduler  | 2025-05-13 01:58:37 [warning  ] Starting call to 
'airflow.sdk.api.client.Client.request', this is the 4th time calling it. 
[airflow.sdk.api.client]
scheduler  | [2025-05-13T01:58:38.897+0000] {_client.py:1026} INFO - HTTP 
Request: PATCH 
http://localhost:8080/execution/task-instances/0196c1ed-1bfe-7616-961a-af093a67ca90/run
 "HTTP/1.1 504 Gateway Timeout"
scheduler  | 2025-05-13 01:58:38 [info     ] Process exited                 
[supervisor] exit_code=<Negsignal.SIGKILL: -9> pid=3129186 signal_sent=SIGKILL
scheduler  | [2025-05-13T01:58:38.907+0000] {local_executor.py:96} ERROR - uhoh
scheduler  | httpx.HTTPStatusError: Server error '504 Gateway Timeout' for url 
'http://localhost:8080/execution/task-instances/0196c1ed-1bfe-7616-961a-af093a67ca90/run'
scheduler  | For more information check: 
https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/504
scheduler  | Correlation-id=0196c75d-8bf5-75ca-bd0b-365e12de83ce
scheduler  | [2025-05-13T01:58:39.917+0000] {scheduler_job_runner.py:1019} 
ERROR - Exception when executing SchedulerJob._run_scheduler_loop
scheduler  | TypeError: HTTPStatusError.__init__() missing 2 required 
keyword-only arguments: 'request' and 'response'
scheduler  | [2025-05-13T01:58:39.921+0000] {local_executor.py:223} INFO - 
Shutting down LocalExecutor; waiting for running tasks to finish.  Signal again 
if you don't want to wait.
scheduler  | [2025-05-13T01:58:39.935+0000] {scheduler_job_runner.py:1031} INFO 
- Exited execute loop
scheduler  | TypeError: HTTPStatusError.__init__() missing 2 required 
keyword-only arguments: 'request' and 'response'
scheduler  | [2025-05-13 01:58:40 +0000] [3129173] [INFO] Shutting down: Master
```
after serveral times of http request:
**HTTP Request: PATCH 
http://localhost:8080/execution/task-instances/0196c1ed-1bfe-7616-961a-af093a67ca90/run
 "HTTP/1.1 504 Gateway Timeout"**, then it would kill scheduler proactively as 
log
```
scheduler  | 2025-05-13 01:58:38 [info] Process exited  [supervisor] 
exit_code=<Negsignal.SIGKILL: -9> pid=3129186 signal_sent=SIGKILL
```
in the host Linux, when I execute 
**wget 
http://localhost:8080/execution/task-instances/0196c1ed-1bfe-7616-961a-af093a67ca90/run**
it prints
```
--2025-05-13 02:02:35--  
http://localhost:8080/execution/task-instances/0196c1ed-1bfe-7616-961a-af093a67ca90/run
Connecting to 10.158.100.9:8080... connected.
Proxy request sent, awaiting response... 504 Gateway Timeout
Retrying.
```
and when I get it in another host with browser 
http://**10.57.200.121**:8080/execution/task-instances/0196c1ed-1bfe-7616-961a-af093a67ca90/run,
 it shows
`{
"detail": "Method Not Allowed"
}`
it's not gateway timeout if I replace localhost by 10.57.200.121.

in airflow 2.10.4, it works well and processes in backend are different with 
processes in airflow 3.0.1rc1, I'm not sure if the difference about process 
communication mechanism causes this issue.


GitHub link: 
https://github.com/apache/airflow/discussions/50382#discussioncomment-13125479

----
This is an automatically sent email for [email protected].
To unsubscribe, please send an email to: [email protected]

Reply via email to