bdeggleston commented on code in PR #3174:
URL: https://github.com/apache/cassandra/pull/3174#discussion_r1548691577
##########
src/java/org/apache/cassandra/tools/nodetool/Repair.java:
##########
@@ -105,6 +105,9 @@ public class Repair extends NodeToolCmd
@Option(title = "paxos-only", name = {"-paxos-only", "--paxos-only"},
description = "If the --paxos-only flag is included, no table data is repaired,
only paxos operations..")
private boolean paxosOnly = false;
+ @Option(title = "accord-only", name = {"-accord-only", "--accord-only"},
description = "If the --accord-only flag is included, no table data is
repaired, only accord operations..")
+ private boolean accordOnly = false;
Review Comment:
I think my objection was exposing the initial iteration of accord repair as
a repair, since it didn't sync data on disk between all nodes. An accord _only_
repair is a little different, it's basically the accord version of a paxos-only
repair. Both paxos and accord only repairs only complete pending operations and
do not sync data.
I'm also don't agree that accord-only repairs wouldn't be useful to end
users. For instance, one of the use cases of paxos repairs is to establish a
lower bound for applied paxos operations so that data can be exported to an
ldap system for processing. Having said that, the accord repair is not on par
with paxos repair since it issues a barrier, but does not wait for it to be
applied on all nodes, and I think that's something I should add before
committing
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
