Before, when searching for an object, we iterated over shards and stopped
right after we found the object. Currently, we need to iterate over all
shards, because, when the object and its GC mark are stored separately, we
could find the object earlier than its GC mark.
Signed-off-by: Aleksey Savchuk <a.savchuk@yadro.com>
This function was very obfuscated. I hope the newer version is more
clear, but IMHO it keeps being bad because:
- Its name is confusing because it checks both the graveyard and the
garbage.
- It has no interface. We use that function in several metabase methods,
it just returns some 'magic' uint8 numbers and has no doc comment, I
mean it's ridiculous.
- It checks out for 'the node being in incorrect state' for some reason
but that result isn't used further. I kept a comment about that but it
has no logic for me.
Signed-off-by: Aleksey Savchuk <a.savchuk@yadro.com>
Currently, when an object on some shard is inhumed, the engine places
a tombstone on the same shard. If the target shard is read-only, the
engine fails.
In that case, we want the engine to correctly place a tombstone on
a different shard, ensuring that:
- An object becomes unavailable if a tombstone was placed on a different
shard. See `TestObjectUnavailableIfTombstoneOnDifferentShard` test.
- GC deletes an object if a tombstone was placed on a different shard.
See `TestObjectDeletedIfTombstoneOnDifferentShard` test.
Signed-off-by: Aleksey Savchuk <a.savchuk@yadro.com>
All error counting and hangling logic is present on the engine level.
Currently, we pass engine metrics with shard ID metric to shard, then
export 3 methods to manipulate these metrics.
In this commits all methods are removed and error counter is tracked on
the engine level exlusively.
Signed-off-by: Evgenii Stratonikov <e.stratonikov@yadro.com>
- `reportShardErrorBackground()` no longer differs from
`reportShardError()`, reflect this in its name;
- reuse common pieces of code to make it simpler.
Signed-off-by: Evgenii Stratonikov <e.stratonikov@yadro.com>
We bind flag that could be specified in config.
This is not a config flag, just a command option.
Also fix TestInitialize failures:
```
Error: Received unexpected error:
number of epochs cannot be less than 1
Test: TestInitialize/16_nodes/force-new-epoch
```
Refs #1372 (945b7c740b)
Signed-off-by: Evgenii Stratonikov <e.stratonikov@yadro.com>
Although these fields could be deleted, I annotated them so that all the
metrics used would be defined in one place.
Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
Target container ID is taken from tombstone: cmd/frostfs-node/object.go:507
Also object of type `TOMBSTONE` contains objectID, so tombstone and
tombstoned object must have the same containerID.
Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
* Remove `relay` field from put streamer as it's no longer used;
* Fix initialization of `Relay` object writer parameter.
Signed-off-by: Airat Arifullin <a.arifullin@yadro.com>
If local EC chunk found, but remote node is off, then `HEAD --raw` request
returns object not found.
Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
These error messages bubble up to human users - adding more context helps
to find the cause of the issue faster.
Signed-off-by: Vitaliy Potyarkin <v.potyarkin@yadro.com>
