octopus
octopus copied to clipboard
Handle slave failures
There is a lot of people requesting a way to handle slave failures.
Some questions:
- What exactly defines a slave as failed?
- In what moment we will mark the slave as down?
- In what moment we will retry the failed slaves?
- The application should boot and warn or raise when a slave is down? (Misconfiguration may not be detected if we allow the application to boot)
I am curious if there was any movement on this. Here are my answers to the questions asked:
- When a no connection error is raised.
- On the first no connection error.
- For me I would say it should be a background process or perhaps not until a server restart.
- Log an error. I don't like the idea of not allowing boot because on server restart that could be a big problem.