Ethan Rose created HDDS-4775:
--------------------------------
Summary: Avoid OM split brain going from 1 node OM to 3 node ratis
without bootstrap
Key: HDDS-4775
URL: https://issues.apache.org/jira/browse/HDDS-4775
Project: Apache Ozone
Issue Type: Sub-task
Reporter: Ethan Rose
Assignee: Ethan Rose
The expected flow to add more OMs to a single node OM ratis cluster is to
adjust the configurations to add the new OMs, and then start the new OMs in
bootstrap mode (supply the -–bootstrap flag on startup) so that they will get
all transaction history from the original OM. However, it is possible for the
user to mistakenly adjust the configs and not start the new OMs with the
--bootstrap command. This will cause the two new OMs to form their own Ratis
ring with a 2/3 majority that can service write requests, while the original OM
is still the leader of a single node Ratis ring and also servicing write
requests. This leads to a split brain scenario that is difficult to detect
without inspecting the logs, because all OMs appear functional and the cluster
is writeable.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
