Scavenge is not freeing up space

Scavenge process is started on a cluster. From the node log I see that the scavenge is working and it says that the new chunks are now smaller than older ones. But in reality all the resulting chunks are 262 MB in size. What should be done to shrink them?

sample from the admin UI scavenge window.

Scavenging chunks 2250 - 2250 complete 102,880,703 00:05:01.3468607 1 chunk(s) scavenged

Scavenging chunks 2249 - 2249 complete 104,370,590 00:04:40.8753761 1 chunk(s) scavenged

Scavenging chunks 2248 - 2248 complete 116,468,944 00:04:12.8976052 1 chunk(s) scavenged

Scavenging chunks 2247 - 2247 complete 108,463,734 00:06:38.5501487 1 chunk(s) scavenged

Scavenging chunks 2246 - 2246 complete 152,424,636 00:06:04.8690420 1 chunk(s) scavenged

Scavenging chunks 2245 - 2245 complete 56,784,595 00:12:42.3637331 1 chunk(s) scavenged

Scavenging chunks 2244 - 2244 complete 154,323,063 00:03:33.1258970 1 chunk(s) scavenged

Scavenging chunks 2243 - 2243 complete 139,157,718 00:04:07.4837172 1 chunk(s) scavenged

``

from the log

[PID:44580:026 2017.03.27 19:56:52.453 TRACE TFChunkScavenger ] SCAVENGING: started to scavenge & merge chunks: #2250-2250 (chunk-002250.000001)

[PID:44580:026 2017.03.27 19:56:52.453 TRACE TFChunkScavenger ] Resulting temp chunk file: 5b9f00cf-da20-44ae-8725-05d1915500d0.scavenge.tmp.

[PID:44580:026 2017.03.27 20:01:50.432 DEBUG TFChunk ] Buffer size is 161280830

[PID:44580:026 2017.03.27 20:01:53.235 INFO TFChunkManager ] Switching chunk #2250-2250 (5b9f00cf-da20-44ae-8725-05d1915500d0.scavenge.tmp)…

[PID:44580:026 2017.03.27 20:01:53.326 INFO TFChunkManager ] File 5b9f00cf-da20-44ae-8725-05d1915500d0.scavenge.tmp will be moved to file chunk-002250.000002

[PID:44580:026 2017.03.27 20:01:53.660 INFO TFChunk ] File chunk-002250.000001 has been marked for delete and will be deleted in TryDestructFileStreams.

[PID:44580:026 2017.03.27 20:01:53.817 INFO TFChunkManager ] Old chunk ##2250-2250 (chunk-002250.000001) is marked for deletion.

[PID:44580:026 2017.03.27 20:01:53.817 TRACE TFChunkScavenger ] Scavenging of chunks:

[PID:44580:026 2017.03.27 20:01:53.817 TRACE TFChunkScavenger ] #2250-2250 (chunk-002250.000001)

[PID:44580:026 2017.03.27 20:01:53.817 TRACE TFChunkScavenger ] completed in 00:05:01.3462736.

[PID:44580:026 2017.03.27 20:01:53.817 TRACE TFChunkScavenger ] New chunk: 5b9f00cf-da20-44ae-8725-05d1915500d0.scavenge.tmp --> #2250-2250 (chunk-002250.000002).

[PID:44580:026 2017.03.27 20:01:53.817 TRACE TFChunkScavenger ] Old chunks total size: 210039425, scavenged chunk size: 107158722.

``

in database directory:

dir | find “2250”

27.03.2017. 23:01 268 439 552 chunk-002250.000002

``

shouldn’t this file be 107158722 bytes in size?

The version of EventStore is 3.9.3, built from source

it can put multiple chunks together resulting in fewer files overall.

How to check if this is the case? For now I am observing that database folder size is constantly growing and I can’t find a way to shrink it.

2017 m. kovas 28 d., antradienis 05:22:33 UTC+3, Greg Young rašė:

Look at scavenge output in UI it will tell you everything that
happened. You can also check the logs of the scavenge but there may be
other things intertwined