You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ignite.apache.org by "Kirill Tkalenko (Jira)" <ji...@apache.org> on 2025/12/10 09:10:00 UTC

[jira] [Created] (IGNITE-27308) Improve the delta file merge pause strategy

Kirill Tkalenko created IGNITE-27308:
----------------------------------------

             Summary: Improve the delta file merge pause strategy
                 Key: IGNITE-27308
                 URL: https://issues.apache.org/jira/browse/IGNITE-27308
             Project: Ignite
          Issue Type: Improvement
            Reporter: Kirill Tkalenko
            Assignee: Kirill Tkalenko
             Fix For: 3.2


Currently, when a checkpoint occurs, we abort the merge of delta files into partition files and then restore it after the checkpoint completes.
If disk writes are slow enough during intensive inserts, we could end up with checkpoints piling up one after another and the merge of delta files never occurs. This could lead to a rapid increase in the number of delta files and an error about too many files being open.

A more flexible approach is needed.
For example, if the number of delta files in a partition is no more than 4, we can abort the merge. If it's more but less than 10, we can slow down the merge slightly. But if it's more, we won't slow down or stop anything.



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