You are viewing a plain text version of this content. The canonical link for it is here.
Posted to hdfs-issues@hadoop.apache.org by "Istvan Fajth (Jira)" <ji...@apache.org> on 2019/08/22 18:03:00 UTC

[jira] [Comment Edited] (HDFS-8178) QJM doesn't move aside stale inprogress edits files

    [ https://issues.apache.org/jira/browse/HDFS-8178?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16913528#comment-16913528 ] 

Istvan Fajth edited comment on HDFS-8178 at 8/22/19 6:02 PM:
-------------------------------------------------------------

I have attached a new patch, that tries to incorporate concerns from [~atm].

The current solution does the following:
 * in JNStorage it uses the FileJournalManager'sĀ purgeLogsOlderThan(long) method to handle old edits removal
 * the purge logic has extended to do the following:
 ** purge edits file segments that are ending before the given transaction ID
 ** purge edits in progress files of any kind (normal leftover, marked as trash, empty, stale, and corrupted that is starting before the given transaction ID
 ** edits in progress files that are normal leftovers but starting after the given transaction ID are marked as stale

Added a test that tests the removal of the new files, and actually as I thought it through now, I will file a new patch shortly with a new test, that tests the removal of marked files.

The suggested patch incorporates HDFS-8304 as well.


was (Author: pifta):
I have attached a new patch, that tries to incorporate concerns from [~atm].

The current solution does the following:
 * in JNStorage it uses the FileJournalManager'sĀ purgeLogsOlderThan(long) method to handle old edits removal
 * the purge logic has extended to do the following:
 ** purge edits file segments that are ending before the given transaction ID
 ** purge edits in progress files of any kind (normal leftover, marked as trash, empty, stale, and corrupted that is starting before the given transaction ID
 ** edits in progress files that are normal leftovers but starting after the given transaction ID are marked as stale

Added a test that tests the removal of the new files, and actually as I thought it through now, I will file a new patch shortly with a new test, that tests the removal of marked files.

> QJM doesn't move aside stale inprogress edits files
> ---------------------------------------------------
>
>                 Key: HDFS-8178
>                 URL: https://issues.apache.org/jira/browse/HDFS-8178
>             Project: Hadoop HDFS
>          Issue Type: Bug
>          Components: qjm
>            Reporter: Zhe Zhang
>            Assignee: Istvan Fajth
>            Priority: Major
>              Labels: BB2015-05-TBR
>         Attachments: HDFS-8178.000.patch, HDFS-8178.002.patch, HDFS-8178.003.patch, HDFS-8178.004.patch, HDFS-8178.005.patch
>
>
> When a QJM crashes, the in-progress edit log file at that time remains in the file system. When the node comes back, it will accept new edit logs and those stale in-progress files are never cleaned up. QJM treats them as regular in-progress edit log files and tries to finalize them, which potentially causes high memory usage. This JIRA aims to move aside those stale edit log files to avoid this scenario.



--
This message was sent by Atlassian Jira
(v8.3.2#803003)

---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-help@hadoop.apache.org