Andrzej Jackowski

@scylladb/scylla-maint pleas consider merging this backport

> [@ptrsmrn](https://github.com/ptrsmrn) [@andrzej-jackowski-scylladb](https://github.com/andrzej-jackowski-scylladb) > > can you look into this audit nemesis, and suggest what's the best way to validate the audit log ? @soyacz, @fruch, as far as I...

> How would you gonna identify duplicates ? > Can we just use python `set()` ? There is `query_num=1M`, I think it is not that much for python to store...

> > > How would you gonna identify duplicates ? > > > > > > Can we just use python `set()` ? There is `query_num=1M`, I think it is...

> BTW, how useful is the QUERY audit if most users are using prepared statements all the time ? doesn't it make sense we'll log the parameters as well ?...

> > I had to dequeue the PR, Andrzej and Marcin identified it to be the root cause of a recent next failure: https://jenkins.scylladb.com/job/scylla-master/job/next/9324/ > > How did we pass...

> @andrzej-jackowski-scylladb I assume this scylladb/scylladb#24658 - right? so it's not even queued, so i have no idea how did it get into Gating No, `audit_test` was introduced in https://github.com/scylladb/scylladb/pull/24231....

@xemul, @asias, I prepared a draft fix for `SEASTAR_ASSERT(size == len);` that creates new `send_ctx` for each call of `posix_datagram_channel::send(const socket_address& dst, packet p)`. But when I repeated my reproduction...

> > Was the posix_datagram_channel designed for such a parallel use? > > Apparently no. But I'd like to see the "draft fix" before discussing why it backtraces :D @xemul...

As mentioned by Avi in https://github.com/scylladb/scylladb/pull/23464#issuecomment-2789229739, implementing then-linked futures here might be problematic: > The problem with then-linked futures is that they can use unbounded memory. It's hard to imagine...