You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "zexin gong (Jira)" <ji...@apache.org> on 2025/02/13 04:26:00 UTC
[jira] [Comment Edited] (FLINK-37313) [BUG] MySQL snapshot read phase is unresponsive for a long time (even up to several hours)
[ https://issues.apache.org/jira/browse/FLINK-37313?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17926654#comment-17926654 ]
zexin gong edited comment on FLINK-37313 at 2/13/25 4:25 AM:
-------------------------------------------------------------
There are two optimization points for this problem:
1. When the high and low watermarks of the snapshot split are the same, binlog reading should be skipped;
2. MySqlStreamingChangeEventSource should record the time for writing to any table so that MySqlBinlogSplitReadTask can correctly verify whether the binlog reading should end.
I'd like to work on a fix for this issue, please assign this issue to me. Thanks!
was (Author: JIRAUSER308676):
There are two optimization points for this problem:
1. When the high and low watermarks of the snapshot split are the same, binlog reading should be skipped;
2. MySqlStreamingChangeEventSource should record the time for writing to any table so that MySqlBinlogSplitReadTask can correctly verify whether the binlog reading is completed.
I'd like to work on a fix for this issue, please assign this issue to me. Thanks!
> [BUG] MySQL snapshot read phase is unresponsive for a long time (even up to several hours)
> ------------------------------------------------------------------------------------------
>
> Key: FLINK-37313
> URL: https://issues.apache.org/jira/browse/FLINK-37313
> Project: Flink
> Issue Type: Bug
> Components: Flink CDC
> Affects Versions: 3.0.0
> Environment: mysql: 8.0.33
> flink-connector-mysql-cdc: 3.3.0
> Reporter: zexin gong
> Priority: Major
> Attachments: 读取binlog.jpg, 读取结果集.jpg
>
>
> The split reading time during the MySQL snapshot phase is abnormal. The specific situation is as follows:
> 1. The {*}snapshot reading is very fast{*}, about a few seconds:
> !读取结果集.jpg!
> 2. The time to *read the high and low watermarks of the snapshot is abnormal* (even up to several hours):
> !读取binlog.jpg!
--
This message was sent by Atlassian Jira
(v8.20.10#820010)