John Spray

some debug unit tests were failing in CI on abandoned failed futures from trying to stat non existent directories: presumably these are tests that destroy their dirs in weird places,...

> on our cloud the amazon health probe just prints this non stop for all clusters, makes all logs useless. Gross, can we fix the client to close its connection...

FAIL test: BasicAuthScaleTest.test_many_users.num_users=500 (1/47 runs) failure at 2022-11-22T15:29:01.731Z:

This is a segfault in SchemaRegistryAutoAuthTest.test_mixed_deletes https://buildkite.com/redpanda/redpanda/builds/19351#0184cf75-d509-42e0-8a45-44285fe849d4 @BenPope I'm tagging onto this issue because it seems like the general vibe is that there's some sort of SR crash buried that...

Here's a segfault in test_post_subjects_subject_versions that I assume is the same underlying issue. FAIL test: SchemaRegistryBasicAuthTest.test_post_subjects_subject_versions (1/55 runs) failure at 2022-12-12T07:10:48.014Z:

> Now, a large compression won't cause reactor stalls, because it can be suspended, but while it is suspended any other task trying to use the compressor (not at all...

> I believe you need to keep the workspace around for the entire [de]compression operation: the context carries context from one ZSTD_compressStream2 call to the next (both settings and [de]compression...

Rebased on the fixes. When they merge, this gets rebased again, then need to eyeball the locations we fixed to make sure this PR doesn't accidentally revert any fixes.

>do i understand correctly that the issue here is that compat_version defaults to version? > if so, it would be sufficient to change the default behavior? i'm not arguing against...

This is good to go