Dynamicpartition sensor crashes due to postgres connection error on Dagster OS. about dagster

[alangenfeld]

Some general tips can be found in this discussion #14771

I assume you are GKE on GCP? Do you see any issues with any of the dagster pods?

How many "stations" / dynamic partitions do you expect to interact with every tick?

all sensors are stuck

Are there multiple sensors structured the same way?

For some context what is happening under the hood is that the dagster daemon is responsible for managing the ticks of sensors/schedules but evaluating function happens in the code server. The daemon should not start the next tick for a sensor until the previous one completes, so I am surprised to see connections pile up unless

• there are multiple of these sensors running concurrently
• the processes are getting OOM killed or something and leaving dangling connections