Hey there. I was working with version 0.14.6 (sad,...
# ask-community
b
Hey there. I was working with version 0.14.6 (sad, I know) and we have just upgraded to 1.17.17. For some reason, I am observing that the button for toggling off sensors and schedules is not working. For example, this sensor was activated and, then, deactivated, and keeps evaluating the necessity for yielding runs. Any ideas on where should I look for debug?
a
some ideas • did you upgrade the instance ie
dagster instance migrate
? • do you possibly have a daemon still running the old version?
b
hi @alex! Thanks for your answer. Our deployment framework is based on Azure's Openshift pods. We downscale them and create new ones with the specified packages. In this context, "migrating" means, for us, to upgrade the requirements.txt that the deployment Dockerfile will use (
RUN pip install --trusted-host [REPO_NAME] --requirement requirements.txt --index-url [URL]
). These requirements are as follows:
azure-identity==1.12.0
holidays==0.13
numpy==1.24.1
pandas==1.5.3
pip==22.3.1
pymongo==4.3.3
azure-core==1.26.2
azure-storage-blob==12.14.1
azure-storage-file-datalake==12.9.1
dagit==1.1.17
dagster==1.1.17
dagster-azure==0.17.17
dagster-databricks==0.17.17
dagster-docker==0.17.17
dagster-graphql==1.1.17
dagster-postgres==0.17.17
dagster-pyspark==0.17.17
dagster-slack==0.17.17
dagster-msteams==0.17.17
dagster-spark==0.17.17
delta-spark==2.2.0
fastparquet==2023.1.0
pyarrow==11.0.0
pyspark==3.3.1
Are we missing any package for the correct setup?
a
packaging looks fine from what i can tell
b
is there something I can do to ensure the migration of all instances?
a
the instance migrate command takes care of updating the database I am not familiar with Azure Openshift but i guess double check that you didn’t miss any pods and that everything that is active is updated to the new packages
b
I did not run
instance migrate command
, as I thought updating the packages would be the same thing. Should I do it explicitly?
Comparing the resulting postgres database with another one we have for another Dagster instance that has been created from scratch in this new version, I see that the instigators table is missing. Do you see a connection? Is this problematic?
there is code in place that should gracefully handle the backcompat for the old table name
jobs
b
We are using Azure Openshift pods, not Kubernetes, so I am not sure that directly applies to our case. I will try to run the
dagster instance migrate
though
It solved the problem! Merely upgrading the packages was not enough.
🎉 1