[ 
https://issues.apache.org/jira/browse/IGNITE-26928?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Aleksandr Polovtsev updated IGNITE-26928:
-----------------------------------------
    Description: 
Here's an exempt from a real use case log (trimmed for clarity):

{noformat}
08:48:45:014 +0000 [INFO][compaction-thread][Compactor] Starting new compaction 
round [compactionId=c2c80c9a-49a5-45c9-8cd0-53a973b51a86, files=1]
08:48:45:016 +0000 [INFO][compaction-thread][Compactor] Compaction round 
finished [compactionId=c2c80c9a-49a5-45c9-8cd0-53a973b51a86, pages=2, 
duration=1ms, avgWriteSpeed=17MB/s]
08:48:45:016 +0000 [INFO][compaction-thread][Compactor] Starting new compaction 
round [compactionId=3c237547-8014-4c48-8148-4089e70d5656, files=1]
08:48:45:021 +0000 [INFO][compaction-thread][Compactor] Compaction round 
finished [compactionId=3c237547-8014-4c48-8148-4089e70d5656, pages=1, 
duration=4ms, avgWriteSpeed=3.18MB/s]
08:48:45:021 +0000 [INFO][compaction-thread][Compactor] Starting new compaction 
round [compactionId=7a4c5b53-6c0d-4474-a6f3-3bdc42ef121e, files=2]
08:48:45:023 +0000 [INFO][compaction-thread][Compactor] Compaction round 
finished [compactionId=7a4c5b53-6c0d-4474-a6f3-3bdc42ef121e, pages=4, 
duration=1ms, avgWriteSpeed=39MB/s]
08:48:45:023 +0000 [INFO][compaction-thread][Compactor] Starting new compaction 
round [compactionId=f9b6017f-9844-4647-a70e-d9a1b15b49a8, files=2]
08:48:45:024 +0000 [INFO][compaction-thread][Compactor] Compaction round 
finished [compactionId=f9b6017f-9844-4647-a70e-d9a1b15b49a8, pages=2, 
duration=1ms, avgWriteSpeed=26MB/s]
{noformat}

There are hundreds of these messages. Since there's no ongoing checkpoint and 
these messages are printed on node startup, compaction is triggered by 
{{PersistentPageMemoryTableStorage#getOrCreatePartitionMetaOnCreatePartition}}. 
We should try triggering compaction less frequently (i.e. at the very end of 
recovery).


> Page Memory storage compaction triggered very often on node start
> -----------------------------------------------------------------
>
>                 Key: IGNITE-26928
>                 URL: https://issues.apache.org/jira/browse/IGNITE-26928
>             Project: Ignite
>          Issue Type: Bug
>            Reporter: Aleksandr Polovtsev
>            Priority: Major
>              Labels: ignite-3
>
> Here's an exempt from a real use case log (trimmed for clarity):
> {noformat}
> 08:48:45:014 +0000 [INFO][compaction-thread][Compactor] Starting new 
> compaction round [compactionId=c2c80c9a-49a5-45c9-8cd0-53a973b51a86, files=1]
> 08:48:45:016 +0000 [INFO][compaction-thread][Compactor] Compaction round 
> finished [compactionId=c2c80c9a-49a5-45c9-8cd0-53a973b51a86, pages=2, 
> duration=1ms, avgWriteSpeed=17MB/s]
> 08:48:45:016 +0000 [INFO][compaction-thread][Compactor] Starting new 
> compaction round [compactionId=3c237547-8014-4c48-8148-4089e70d5656, files=1]
> 08:48:45:021 +0000 [INFO][compaction-thread][Compactor] Compaction round 
> finished [compactionId=3c237547-8014-4c48-8148-4089e70d5656, pages=1, 
> duration=4ms, avgWriteSpeed=3.18MB/s]
> 08:48:45:021 +0000 [INFO][compaction-thread][Compactor] Starting new 
> compaction round [compactionId=7a4c5b53-6c0d-4474-a6f3-3bdc42ef121e, files=2]
> 08:48:45:023 +0000 [INFO][compaction-thread][Compactor] Compaction round 
> finished [compactionId=7a4c5b53-6c0d-4474-a6f3-3bdc42ef121e, pages=4, 
> duration=1ms, avgWriteSpeed=39MB/s]
> 08:48:45:023 +0000 [INFO][compaction-thread][Compactor] Starting new 
> compaction round [compactionId=f9b6017f-9844-4647-a70e-d9a1b15b49a8, files=2]
> 08:48:45:024 +0000 [INFO][compaction-thread][Compactor] Compaction round 
> finished [compactionId=f9b6017f-9844-4647-a70e-d9a1b15b49a8, pages=2, 
> duration=1ms, avgWriteSpeed=26MB/s]
> {noformat}
> There are hundreds of these messages. Since there's no ongoing checkpoint and 
> these messages are printed on node startup, compaction is triggered by 
> {{PersistentPageMemoryTableStorage#getOrCreatePartitionMetaOnCreatePartition}}.
>  We should try triggering compaction less frequently (i.e. at the very end of 
> recovery).



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to