Israel Fruchter
Israel Fruchter
Those usually mean we were sending out too much logs, I.e. Scylla was generating too much logs to be sent via network. SCT needs the log at runtime where the...
> @soyacz > > > Possibly https://github.com/scylladb/scylla-cluster-tests/pull/6657 introduced regression or this test is somehow specific that led to this error: syslog-ng[4612]: EOF occurred while idle; fd='37' > > I have...
> > I was thinking about it, and possible reason is to have time synchronized for multi-dc scenarios > > Multi-DC makes things even worse for client-side timestamps, the timestamps...
@soyacz can you raise the issues from above, and put them on our board ?
> Another occurrence of the issue, which introduced false positive in the test result during [longevity-schema-topology-changes-12h-test](https://argus.scylladb.com/test/e0d56457-6e58-49db-9f1e-ab59bfeab88a/runs?additionalRuns[]=a507ebff-0f45-4dd0-85a8-a0ac714c9bae). > The test issues `nodetool rebuild` command during AddRemoveDc Nemesis and then waits for...
@aleksbykov please look into this one
> > IMHO, `1 hour` for decommission doesn't sound as something "too much" for a healthy busy DB cluster. Do we have somewhere the limits for the decommission operation? >...
@amnonh why this isn't part of scylla itself ? some as we do node_exporter ?, i.e. user installing from RPMs/DEBs won't need this ?
the only problem with this one, is if the test was using this configuration, it wouldn't append to it as one might expect. we'll need to get it figured out,...
> maybe we could introduce `scylla_logger_settings` param? This should be appendable to `append_scylla_args` logging is just an example, there could be more things one might want to enable from the...