Fix user object counter #911

dstepanov-yadro · 2024-01-15T15:07:41Z

dstepanov-yadro commented

2024-01-15 15:07:41 +00:00

Relates #904

The bug was caused by two problems:

The object was saved on node 1, but the metric was not updated. The fact of saving on node 1 is confirmed by logs. Perhaps the problem is updating the metrics. I couldn't reproduce it, but just in case, I added additional logging for Inhume operation.
The object was simultaneously written to node 3 by the put process and the replicator. Because of this, the counter was increased twice. I fixed that.

Relates #904 The bug was caused by two problems: 1. The object was saved on node 1, but the metric was not updated. The fact of saving on node 1 is confirmed by logs. Perhaps the problem is updating the metrics. I couldn't reproduce it, but just in case, I added additional logging for `Inhume` operation. 2. The object was simultaneously written to node 3 by the `put` process and the replicator. Because of this, the counter was increased twice. I fixed that.

dstepanov-yadro changed title from ~~fix/user_counter~~ to Fix user object counter

2024-01-15 15:08:01 +00:00

dstepanov-yadro requested review from storage-core-committers 2024-01-15 15:08:07 +00:00

dstepanov-yadro requested review from storage-core-developers 2024-01-15 15:08:19 +00:00

acid-ant approved these changes 2024-01-16 07:40:57 +00:00

aarifullin approved these changes 2024-01-16 07:41:37 +00:00

fyrchik reviewed 2024-01-16 13:47:06 +00:00

pkg/local_object_storage/engine/error_test.go Outdated

					
				@ -221,6 +221,7 @@ func TestBlobstorFailback(t *testing.T) {

					checkShardState(t, te.ng, te.shards[0].id, 0, mode.ReadWrite)

					checkShardState(t, te.ng, te.shards[1].id, 0, mode.ReadWrite)

					require.NoError(t, te.ng.Close(context.Background()))

fyrchik commented

2024-01-16 12:04:51 +00:00

Why not do it in defer, similar to other changes in the recent refactoring?

dstepanov-yadro commented

2024-01-16 19:18:30 +00:00

There are two te instances created, first instance is closed by regular function call, so I decided to do the same way with the second instance.

There are two `te` instances created, first instance is closed by regular function call, so I decided to do the same way with the second instance.

fyrchik marked this conversation as resolved

pkg/local_object_storage/metabase/put.go Outdated

					
				@ -87,2 +89,3 @@

					err = db.boltDB.Batch(func(tx *bbolt.Tx) error {

						return db.put(tx, prm.obj, prm.id, nil, currEpoch)

						var e error

						res, e = db.put(tx, prm.obj, prm.id, nil, currEpoch)