[docdb] Master falling behind does not get automatically remote bootstrapped #4799
Labels
area/docdb
YugabyteDB core features
kind/enhancement
This is an enhancement of an existing feature
priority/medium
Medium priority issue
Projects
Jira Link: DB-1842
Currently, if a master falls behind to the point that the leader no longer has logs, or its follower unavailability window passes (default 15m for both), it would require manual intervention to remove it from the quorum and add it back.
At the raft level, we have this flag to control w/e or nor to evict failed followers:
evict_failed_followers
. This is disabled for masters though, as there is no automated mechanism for them to be added back into the quorum!Moreover, until #2007 is resolved, the current steps of issuing a remove, followed by an add, require that in between, we actually wipe the master data and start back the master in shell mode (empty
master_addresses
flag).Ideally, we'd have some mechanic through which we would automatically trigger a remote bootstrap for the master, if it comes back after the default 15m window.
cc @iSignal @hectorgcr
The text was updated successfully, but these errors were encountered: