[
https://issues.apache.org/jira/browse/LUCENE-9968?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17350447#comment-17350447
]
Michael McCandless commented on LUCENE-9968:
--------------------------------------------
OK, hmm, a couple ideas:
If I remember correctly, that parameter {{forceMergeMBPerSec}} takes effect
only on the written bytes, not the bytes being read, so you would also see 10
MB/sec of bytes being read (for the old segments that are being merged), and
possibly even higher MB/sec if those segments have many deletions, for example.
Second, the iostat output you showed above is an "instantaneous" measure of how
many bytes the OS just flushed from its write cache to the underling storage
device. Many lines are 0, then suddenly there is a burst of e.g. 30 MB or 100
MB. You would need to average those rates over a longer period to see what
actual MB/sec on average you see, and from the numbers above, it looks like it
might actually be less than 10 MB/sec on average (assuming your iostat is
printing once per second?).
> Change forceMergeMBPerSec to 10 but iostat show bigger than 10MB/s
> ------------------------------------------------------------------
>
> Key: LUCENE-9968
> URL: https://issues.apache.org/jira/browse/LUCENE-9968
> Project: Lucene - Core
> Issue Type: Bug
> Components: core/index
> Affects Versions: 6.5.1
> Reporter: jiarun.yang
> Priority: Major
> Attachments: image-2021-05-22-18-13-48-143.png,
> image-2021-05-22-18-16-29-513.png, image-2021-05-22-18-17-21-431.png
>
>
> !image-2021-05-22-18-17-21-431.png! I changed the param
> forceMergeMBPerSec，then to forcemerge, I find io wkB/s is bigger than 10mb/s
> by iostat
> !image-2021-05-22-18-16-29-513.png!!image-2021-05-22-18-16-53-421.png!
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org
