Piotr Smaron
Piotr Smaron
Version 9: - worked around the issue where we cannot perform ALTER for a single-DC setup. The workaround may result in tablets imbalance, the real issue needs to be yet...
@raphaelsc @denesb could you please take a look at this PR and help me resolve some of the TODO comments? I'll also appreciate any other comments (but let's leave the...
Version 10: - split the PR into 2: here we'll deal only with the processing path for the alter tablets-enabled keyspace, while here https://github.com/scylladb/scylladb/pull/17116 is the method that calculates new...
Version 10: - the production code mostly works, but there are known bugs, like failures when decreasing RF for some cases, and the tablets tests fail because of that -...
Version 11: - apparently I had some trash cached and needed to fix the build. - Also, because of the above, the code was not working - now it is...
> @ptrsmrn - seems related to service levels - which are moving to Raft soon, no? @mykaul yes, @Jadw1 is working on it.
The title is a bit misleading, I'd just write in plain words, that we either add or remove tablets' replicas when ALTERing tablets KS.
I remember you also had a UT for this change, can you add it here as well?
> > I remember you also had a UT for this change, can you add it here as well? > > This function is only available from within this TU....
I think we're missing a case when someone provides generic `replication_factor` tag in ALTER KS statement and only handle a case where every DC with its RF is stated explicitly....