Add writecache inhume tests #1028

Merged

fyrchik merged 1 commit from dstepanov-yadro/frostfs-node:fix/writecache_consistency into master

2024-09-04 19:51:07 +00:00

dstepanov-yadro commented

2024-03-06 09:49:23 +00:00

Member

Closes #634

See issue for details.

Delete method for writecache called only by shard.Inhume, so there are two possible ways:

object was flushed to blobstore before shard.Inhume, so GC will eventually delete this object from metabase and blobstore
object was deleted from writecache during shard.Inhume, then GC will eventually delete this object from metabase only
GC will skip object from deletion if writecache stores this object.

I have added unit tests for this.

Writecache has specific usage, so objectID is hash of object's payload and headers. So having the same objectID with different payloads is almost impossible.

Closes #634 See issue for details. 1. `Delete` method for writecache called only by `shard.Inhume`, so there are two possible ways: - object was flushed to blobstore before `shard.Inhume`, so GC will eventually delete this object from metabase and blobstore - object was deleted from writecache during `shard.Inhume`, then GC will eventually delete this object from metabase only GC will skip object from deletion if writecache stores this object. I have added unit tests for this. 2. Writecache has specific usage, so objectID is hash of object's payload and headers. So having the same objectID with different payloads is almost impossible.