Xandra is unable to recover from `DBConnection.ConnectionPool` crash
Closed this issue · 1 comments
harunzengin commented
We have chaos-monkey style tests where we use havoc in order to test the resiliency of our process tree. It looks like when the DBConnection.ConnectionPool is killed, we still have the old pool_pid in %Xandra.Cluster.Pool{} state.
I'm not sure whether we need to fix this and how likely it is for the ConnectionPool to die and get restarted in production, but we have to explicitly make sure that DBConnection.ConnectionPool processes don't get killed in tests.
whatyouhide commented
Nice catch, didn't even know about havoc. Love it. I added some tests and fixed this in the commit above ☝️