1. Change the default HA mode not to be "minority" mode [Fixes the quorum problem]
2. Modify the Postgres configuration to use synchronous replication.
Need to handle
1. A patch on top of 4.6 [using patchify]
2. Document upgrade flow for 4.6 users, earlier version users, and from scratch 4.6 deployment
from the description above:
“2. Document upgrade flow for 4.6 users, earlier version users, and from scratch 4.6 deployment”
This should only b handled for 4.6 (either on clean install or running 4.6 that just got this patch through patchify)
earlier version users should upgrade to 4.6
The point about earlier versions is that we are clear about the path. e.g. if I have 4.4, I should (for example, I am not sure if this is the true procedure)
Move to 4.6 clean version by snapshot+restore path on 4.6
Apply the latest 4.6 patches
re-create the cluster
IIUC this is the exact procedure , Lukasz please respond in case this is not the cae.
The patch was implemented in https://github.com/cloudify-cosmo/patchify/pull/128
However note that for applying the patch to an existing cluster, additional commands must be executed, the summary is in a comment under the PR.
As part of this change, not only was the db made to use synchronous replication, but also several smaller changes were introduced, such as:
after 3 minutes of failing the db health check, the manager will attempt to follow again
This is due to additional issues found in testing