Kamil Braun comments

Results 568 comments of


                                            Kamil Braun

Timing problem when retrieve CDC streams: runtime_error (Could not find CDC generation in distributed system tables)

@mykaul in raft-topology mode CDC generations are distributed differently, through group 0 instead of a regular replicated table. I think there's little point in pursuing this issue. If the other...

Jepsen tests started to be flaky

> was OK on 2024.1.12 but earlier you said > The issue has reproduced again in 2024.12 - [jepsen-test-all](https://argus.scylladb.com/tests/scylla-cluster-tests/16cd94ba-45bc-4a9b-b648-f96d64202910). ?

test_create_lots_of_schema_churn_with_node_down failure: Scylla interrupts its own Raft snapshot transfers when a node is restarted

node 2 restarted which triggered raft snapshot transfer from node 1 (raft leader) this took a long time because the snapshot was large (lots of schema) the RPC was interrupted...

test_create_lots_of_schema_churn_with_node_down failure: Scylla interrupts its own Raft snapshot transfers when a node is restarted

Faster to just send a fix: https://github.com/scylladb/scylladb/pull/20789

test_create_lots_of_schema_churn_with_node_down failure: Scylla interrupts its own Raft snapshot transfers when a node is restarted

@d-helios why did you close the issue? I'm reopening. The errors are harmless as I explained here: https://github.com/scylladb/scylladb/issues/20588#issuecomment-2363492593 but it's still an issue that they're being reported. But a low...

test_create_lots_of_schema_churn_with_node_down failure: Scylla interrupts its own Raft snapshot transfers when a node is restarted

increasing priority due to frequency

test_create_lots_of_schema_churn_with_node_down failure: Scylla interrupts its own Raft snapshot transfers when a node is restarted

I sent https://github.com/scylladb/scylla-dtest/pull/5284 as a workaround. Will reduce back to P3 once the workaround is in.

test_create_lots_of_schema_churn_with_node_down failure: Scylla interrupts its own Raft snapshot transfers when a node is restarted

scylladb/scylla-dtest@8a518e5a7326e300b0dd7074533ac57eb0fef279 contains a workaround, the test should no longer be flaky. (That last run didn't have the commit yet) Reducing priority to P3.

test_create_lots_of_schema_churn_with_node_down failure: Scylla interrupts its own Raft snapshot transfers when a node is restarted

I renamed the issue to mention the root cause and added explanation to the original post.

test_create_lots_of_schema_churn_with_node_down failure: Scylla interrupts its own Raft snapshot transfers when a node is restarted

@lsfreitas @mykaul why are we looking at 2 month old jenkins runs? That run is from Oct 13; the workaround got merged at Nov 6.