Ashwani Raina has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/20546 )

Change subject: [compaction] Skip memory allocation for ancient undo deltas
......................................................................


Patch Set 4:

(4 comments)

http://gerrit.cloudera.org:8080/#/c/20546/4//COMMIT_MSG
Commit Message:

http://gerrit.cloudera.org:8080/#/c/20546/4//COMMIT_MSG@10
PS4, Line 10: we
> it
I guess "compaction code doesn't" would sound better.


http://gerrit.cloudera.org:8080/#/c/20546/4//COMMIT_MSG@11
PS4, Line 11: ancient
> Please give more detail about the definition `ancient`?
I can give a little more detail as part of response to your comment. It 
wouldn't be worth adding that detail in the commit message.

There is a flag (tablet_history_max_age_sec) that denotes the duration for 
which UNDO deltas are retained. If the age of UNDO deltas are more than this 
flag value, it is considered to be ancient and is eventually deleted either via 
GC maintenance op or rowset compaction maintenance op.

The use case of retaining these deltas is to cater to  historical scan 
requests, incremental backups, etc. Default value for 
tablet_history_max_age_sec is a week i.e. any historical scan back upto a week 
can be served and any scan request for data older than one week will only serve 
the oldest delta present i.e. 7 days old. One could change the flag value based 
upon their requirement of how old a data one wants to retain.


http://gerrit.cloudera.org:8080/#/c/20546/4/src/kudu/tablet/compaction.cc
File src/kudu/tablet/compaction.cc:

http://gerrit.cloudera.org:8080/#/c/20546/4/src/kudu/tablet/compaction.cc@1243
PS4, Line 1243: in the calle
> Can you describe the caller more clear?
Done.


http://gerrit.cloudera.org:8080/#/c/20546/4/src/kudu/tablet/compaction.cc@1307
PS4, Line 1307:         //    in caller while processing ancient undo deltas. 
Essentially,
> Maybe you can remove this comments to line 1282.
The comment is relevant and desirable here because this provides details about 
how the workflow of "INSERT after DELETE" and "DELETE being the last mutation" 
work as far as compaction is concerned.

Even though there is no code change here, the comments explain the existing 
code for posterity.



-- 
To view, visit http://gerrit.cloudera.org:8080/20546
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: Ibb41636a3ac063478fe181560d4ec85dadeb0ef3
Gerrit-Change-Number: 20546
Gerrit-PatchSet: 4
Gerrit-Owner: Ashwani Raina <ara...@cloudera.com>
Gerrit-Reviewer: Alexey Serbin <ale...@apache.org>
Gerrit-Reviewer: Ashwani Raina <ara...@cloudera.com>
Gerrit-Reviewer: Kudu Jenkins (120)
Gerrit-Reviewer: Tidy Bot (241)
Gerrit-Reviewer: Wang Xixu <1450306...@qq.com>
Gerrit-Reviewer: Zoltan Martonka <zmarto...@cloudera.com>
Gerrit-Comment-Date: Fri, 27 Oct 2023 08:25:38 +0000
Gerrit-HasComments: Yes

Reply via email to