And summarizing the changes we've put in place here.
1. We started doing our juju bootstrap --to a VM on a node which isn't the PostgreSQL master. This didn't stop the problem from occurring frequently.
2. We have a bcache on /var in write-back mode. This seems to have made the issue stop occurring.
Did the lru change Blake described get committed?
And summarizing the changes we've put in place here.
1. We started doing our juju bootstrap --to a VM on a node which isn't the PostgreSQL master. This didn't stop the problem from occurring frequently.
2. We have a bcache on /var in write-back mode. This seems to have made the issue stop occurring.
Did the lru change Blake described get committed?