astacus
astacus copied to clipboard
Aiven-Open
Clustered database backup
CPU intensive steps running in the main thread block all async operations. This include other concurrent requests, leaving astacus completely unresponsive. Start moving some steps entirely in a thread pool...
When restoring a cluster, the new nodes AZ distribution may not match the one stored in the backup manifest. In this case, the `MapNode` fails even though the new nodes...
It is not sufficient to use one set of dev dependencies. In particular, pytest-asyncio has introduced a breaking change in f39.
Presently the patterns we use to pick the files that go into the backup for Cassandra do not start at the keyspace level; they include the data directory `data`: ```...
The sqlite approach should be probably adapted to it too - keep track of the files that are in use there, instead of keeping (millions of) Python strings in memory....
The main change here is that we need to work for both pydantic 1.10.17 and pydantic 2.
# (DRAFT) About this change - What it does Replace delete_item with delete_items in order to support bulk deletion of object storage files for gcp. Reference - https://github.com/Aiven-Open/rohmu/pull/193
Run the restore integration suite both with direct and local copy.
https://github.com/ClickHouse/ClickHouse/pull/56132. Cache the output of get_clickhouse_version if we are going to call it a lot of times. With this, all the Astacus tests pass for ClickHouse 24.3.