-
|
Similar issue: #33744 (comment) I've deployed Airflow on GKE with the I shelled into the airflow@airflow-scheduler-7bb8dc9dd7-27thc:/opt/airflow$ pip freeze | grep redis
apache-airflow-providers-redis==3.4.0
google-cloud-redis==2.13.2
redis==4.6.0
airflow@airflow-scheduler-7bb8dc9dd7-27thc:/opt/airflow$ pip freeze | grep celery
apache-airflow-providers-celery==3.4.1
celery==5.3.4
airflow@airflow-scheduler-7bb8dc9dd7-27thc:/opt/airflow$ python
Python 3.8.18 (default, Nov 1 2023, 14:44:54)
[GCC 10.2.1 20210110] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import redis
>>> from redis import client
>>>
|
Beta Was this translation helpful? Give feedback.
Replies: 3 comments 9 replies
-
|
Wrong POD. Mostly something on your celery workers causes redis to be upgraded. But what it is and how - impossible to say . You need to check what your |
Beta Was this translation helpful? Give feedback.
-
|
Both |
Beta Was this translation helpful? Give feedback.
-
|
Likely you have something that makes your celery image to upgrade redis. Check it on running celery worker. No-one knows what you do there - until you check and see what happens there. |
Beta Was this translation helpful? Give feedback.
-
|
Also, it's likely that you have mistake in your deployment and either old containers or ones that use different images. Look for typos and the like. |
Beta Was this translation helpful? Give feedback.
-
|
This is probably the answer. I've updated our helm charts to a different image ( |
Beta Was this translation helpful? Give feedback.
-
|
seeing same exact error in apache/airflow:2.7.3-python3.9 too. I was using apache/airflow:2.7.3-python3.8 as well and after seeing this discussion upgraded to apache/airflow:2.7.3-python3.9 and it worked for sometime, now I see it again. Any workaround for this ? |
Beta Was this translation helpful? Give feedback.
-
|
I do not know the reason, so no, it would be great to know WHAT you have changed (you had to change something - it could not just start happening - some of your deployment, customisations etc. must have done something. If we know what that SOMETHING was, we might attempt to diagnose and fix it, but so far we have no idea. |
Beta Was this translation helpful? Give feedback.
-
|
New software Installation, building image automatically, clearing caches, pulling images, etc. etc. - if we know what could have triggered it, then we can do something to check it - any information, explanation, details what happened would be helpful. |
Beta Was this translation helpful? Give feedback.
-
|
in Helm, I set workers.persistence.enabled=false since I read logs from elasticsearch but I see that there is a specific logs.persistence field as well. Then what's difference between these two persistency ? When I set workers.persistence.enabled=true earlier, it was using Statefulset and auto-provisioned PVC name was logs-airflow-worker-0. When I set workers.persistence.enabled=false, it turned from Statefulset to Deployment. Don't know if it's related or not |
Beta Was this translation helpful? Give feedback.
-
|
Nevertheless, reverting back changes doesn't help I still get AttributeError: module 'redis' has no attribute 'client' although see this |
Beta Was this translation helpful? Give feedback.
-
|
My guess is that SOMEWHERE on your PYTHONPATH you have There is a whole documentation about that in Airflow docs that you should read and understand how python module importing work and how you should avoid the overlaps of generic names: https://airflow.apache.org/docs/apache-airflow/stable/administration-and-deployment/modules_management.html If you also want a bit more precise instructions and warnings of what you should learn and avoid, here is an update to the documentaiton that you should read (it is scheduled to be released together with 2.8.2) https://github.com/apache/airflow/pull/36927/files Please read and digest those docs, and it would be great, when you find the cause you come back here and explain what went wrong in your deployment so that others could also see it and avoid it. |
Beta Was this translation helpful? Give feedback.
-
|
For those still running into this, a detailed explanation of the cause can be found in #41359 |
Beta Was this translation helpful? Give feedback.
Also, it's likely that you have mistake in your deployment and either old containers or ones that use different images. Look for typos and the like.