[ https://issues.apache.org/jira/browse/CASSANDRA-12106?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17435635#comment-17435635 ]
Josh McKenzie commented on CASSANDRA-12106: ------------------------------------------- Ok, quite a few commits back and forth in on that PR and it's pretty tidied up. I'll pull a comment I left there here as it's the one outstanding problem with the design we're not going to fix on this initial commit (it's a tradeoff): {quote}The last outstanding thing I can't think of a good solution to is what to do if a node is gc thrashing, read and write CQL stages are a mess, and denylisting requires CQL mutation to work. While we _could_ append a new deny listed key to the DenylistEntry's in the cache independently of, or before, the CQL mutation that adds it to the dedicated store for the denylist, that means we'll likely have a disjoint between what's deny listed in our cache vs. what's in CQL and makes the in-memory cache the authority vs. what's in the DB. I'm actually somewhat receptive to that, since the JMX call queries what's in the cache and actively denylisting, but I think we should probably leave that up to a follow-up ticket rather than trying to tackle that on the tail end of this... marathon. {quote} The tests are kind of a mess: [JDK 8|https://app.circleci.com/pipelines/github/josh-mckenzie/cassandra/134/workflows/2cd807fb-2393-4fd5-89c8-aa5c71eef491] [JDK 11|https://app.circleci.com/pipelines/github/josh-mckenzie/cassandra/134/workflows/3c36936f-97c7-4d08-9383-e9afc476fd36] Here's where I've gotten on debugging the failing test front: * j8 + 11 dtest no vnodes: test_bootstrap_with_reset_bootstrap_state: Created CASSANDRA-17076; happening on trunk * j8 + 11 dtest with vnodes: test_bootstrap_binary_disabled: Created CASSANDRA-17077; happening on trunk * j8 unit: testPersistentStatistics: Created CASSANDRA-17078; can't reproduce locally on branch or trunk * j8 dtest novnode: ** test_rolling_upgrade_with_internode_ssl x3 ** test_rolling_upgrade x2: Created CASSANDRA-17079; also happening on trunk ** test_drop_compact_storage_mixed_cluster: Created CASSANDRA-17080 ** test_bootstrap_with_reset_bootstrap_state: Created CASSANDRA-17081 * j8 jvm upgrade: InvocationTargetException - being fixed in CASSANDRA-17050 * j11 unit: testNoTreesRetainedAfterDifference - happening on trunk, tracked in CASSANDRA-17039 I'm going to give it a couple days for us to talk through what's going on w/the testing situation, maybe see CASSANDRA-17050 cleared out, before merging this in. I also want to read through {{MigrationManager.evolveSystemKeyspace}} again and think about the behavior in mixed version cluster states for generation changes in SystemDistributed (have some other changes queued for that in the auth domain as well). > Add ability to blocklist / denylist a CQL partition so all requests are > ignored > ------------------------------------------------------------------------------- > > Key: CASSANDRA-12106 > URL: https://issues.apache.org/jira/browse/CASSANDRA-12106 > Project: Cassandra > Issue Type: New Feature > Components: Legacy/Local Write-Read Paths, Local/Config > Reporter: Geoffrey Yu > Assignee: Josh McKenzie > Priority: Low > Fix For: 4.x > > Attachments: 12106-trunk.txt > > > Sometimes reads/writes to a given partition may cause problems due to the > data present. It would be useful to have a manual way to blocklist / denylist > such partitions so all read and write requests to them are rejected. -- This message was sent by Atlassian Jira (v8.3.4#803005) --------------------------------------------------------------------- To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org For additional commands, e-mail: commits-h...@cassandra.apache.org