Piotr Smaron
Piotr Smaron
> > 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...
We have to eliminate the query timeout when ALTERing KS. It probably doesn't matter if it's tablet-enabled KS or not, both can have the timeout disabled, which simplifies the implementation....
Closing per https://github.com/scylladb/scylladb/issues/18931#issuecomment-2210575434
> @ptrsmrn - is the move to Raft helping in this use case? (or are we still stuck with this LOCAL_ONE queries, regardless?) @mykaul sorry, I was sure I already...
auth v2 (working under RAFT) is on master, but per @nuivall it's better to wait until https://github.com/scylladb/scylladb/issues/17738 land on master too.
Works for me locally. This code: ``` auto ks_md_update = _attrs->as_ks_metadata_update(ks_md, tm, feat); auto serialized_ks_md_update = fmt::format("{}", ks_md_update); mylogger.warn("ks_md_update: {}", ks_md_update); mylogger.warn("serialized_ks_md_update: {}", serialized_ks_md_update); fmt::format_parse_context ctx{serialized_ks_md_update}; auto deserialized_ks_md_update = fmt::formatter::parse(ctx);...
@mykaul did you mean me when assigning this item? This is rather tablets realm.
UPDATE: - rebased on top of the latest master to resolve test_tablets.py conflicts, - addressed @bhalevy 's comments.
UPDATE: restarted the CI.
UPDATE: re-kicked CI passed, which means previous faliures were due to infra congestion.