[
https://issues.apache.org/jira/browse/KAFKA-15046?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17755904#comment-17755904
]
Haruki Okada commented on KAFKA-15046:
--
I submitted a patch
[
https://issues.apache.org/jira/browse/KAFKA-15046?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17755898#comment-17755898
]
Haruki Okada commented on KAFKA-15046:
--
After dug into the fsync call paths in detail, I summarized
[
https://issues.apache.org/jira/browse/KAFKA-15046?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17755351#comment-17755351
]
Haruki Okada commented on KAFKA-15046:
--
[~junrao] Hi, sorry for the late response.
Thanks for your
[
https://issues.apache.org/jira/browse/KAFKA-15046?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17731375#comment-17731375
]
Jun Rao commented on KAFKA-15046:
-
The potential issue with moving the LeaderEpochFile flushing to an
[
https://issues.apache.org/jira/browse/KAFKA-15046?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17731119#comment-17731119
]
Jun Rao commented on KAFKA-15046:
-
[~ocadaruma] : Thanks for identifying the problem. Another way to
[
https://issues.apache.org/jira/browse/KAFKA-15046?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17729618#comment-17729618
]
Luke Chen commented on KAFKA-15046:
---
[~ocadaruma] , thanks for the help!
Ping me when PR is created.
[
https://issues.apache.org/jira/browse/KAFKA-15046?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17729561#comment-17729561
]
Haruki Okada commented on KAFKA-15046:
--
I see, thank you for pointing out.
Hmm, now I agree with
[
https://issues.apache.org/jira/browse/KAFKA-15046?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17728606#comment-17728606
]
Luke Chen commented on KAFKA-15046:
---
Currently, the RW lock in LeaderEpochFileCache is to protect the
[
https://issues.apache.org/jira/browse/KAFKA-15046?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17728551#comment-17728551
]
Haruki Okada commented on KAFKA-15046:
--
[~showuon] Maybe I linked wrong file.
What I thought is
[
https://issues.apache.org/jira/browse/KAFKA-15046?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17728547#comment-17728547
]
Luke Chen commented on KAFKA-15046:
---
I agree that io_uring is a good way to fix that, but it needs
[
https://issues.apache.org/jira/browse/KAFKA-15046?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17728515#comment-17728515
]
Haruki Okada commented on KAFKA-15046:
--
Yeah, io_uring is promising.
However it only works with
[
https://issues.apache.org/jira/browse/KAFKA-15046?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17728275#comment-17728275
]
Divij Vaidya commented on KAFKA-15046:
--
Yes that is right, leaderEpochCheckpoint is another I/O
[
https://issues.apache.org/jira/browse/KAFKA-15046?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17728266#comment-17728266
]
Haruki Okada commented on KAFKA-15046:
--
Hm, when I dug into further this, I noticed there's another
[
https://issues.apache.org/jira/browse/KAFKA-15046?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17728262#comment-17728262
]
Haruki Okada commented on KAFKA-15046:
--
Oh I haven't noticed there's another ticket and already the
[
https://issues.apache.org/jira/browse/KAFKA-15046?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17728260#comment-17728260
]
Divij Vaidya commented on KAFKA-15046:
--
Thank you for the investigation folks. We have an active PR
[
https://issues.apache.org/jira/browse/KAFKA-15046?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17728214#comment-17728214
]
Haruki Okada commented on KAFKA-15046:
--
If the suggestion (stop fsync-ing) makes sense, I'm happy
16 matches
Mail list logo