maas takes a while to recover after primary database is moved

[Jason Hobbs]

This is with 2.5.2-7523-ge4ecbd54d-0ubuntu1~18.04.1.

We have a test where we kill the postgres master, wait for the failover to happen, then try to use maas.

We wait 75 seconds between killing the database master and trying to use maas.

However, we still get an error back from MAAS occasionally:

"500 Internal Server Error (SSL SYSCALL error: EOF detected"

[Andres Rodriguez (andreserl)]

Marking this as incomplete provided that:

1. There are no regiond.log's attached for any of the regions.
2. The error above is not clear where it comes from, it talks about 'SSL SYSCALL' when MAAS doesn't even support nor configures SSL.
3. My understanding is that the configuration for postgresql HA has recently changed. Is there any indication that this could be the result of that ?
4. There's no information about the state of postgresql when you see that error. Can you confirm it recover successfully and the database is up to date and no errors with postgresql?

[Björn Tillenius (bjornt)]

We did some changes recently in how MAAS detects that a database connection is broken and need to be reconnected. Do you see the same behavior with MAAS 2.6.0?

[Launchpad Janitor (janitor)]

[Expired for MAAS because there has been no activity for 60 days.]