You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I'm wondering what is the recommended rollout strategy. For example, I work with DigitalOcean's PaaS. When I do a new deployment, the system won't necessarily wait until my worker gracefully shuts down before sending it a SIGKILL.
As a result I often end up with stalled tasks that were interrupted while running.
A way that I can imaging solving this is by having something like this:
A command tells all running workers to stop starting tasks
Another command waits until no tasks are doing anymore (the command would run as a pre-deploy command and can delay the deployment process as needed)
Once all tasks are done, then the PaaS will SIGTERM the worker(s)
And finally a new worker process is deployed
This is how I imagine things. I'm not sure what is the "canonical" way to deal with this problem?
Thanks!
The text was updated successfully, but these errors were encountered:
Hi,
I'm wondering what is the recommended rollout strategy. For example, I work with DigitalOcean's PaaS. When I do a new deployment, the system won't necessarily wait until my worker gracefully shuts down before sending it a SIGKILL.
As a result I often end up with stalled tasks that were interrupted while running.
A way that I can imaging solving this is by having something like this:
This is how I imagine things. I'm not sure what is the "canonical" way to deal with this problem?
Thanks!
The text was updated successfully, but these errors were encountered: