You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@bookkeeper.apache.org by Matteo Merli <mm...@yahoo-inc.com> on 2015/04/10 23:45:25 UTC
Review Request 33094: Use nanoseconds to calculate poll timeout when
doing group commit
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/33094/
-----------------------------------------------------------
Review request for bookkeeper.
Bugs: BOOKKEEPER-850
https://issues.apache.org/jira/browse/BOOKKEEPER-850
Repository: bookkeeper-git
Description
-------
When using group commit to reduce the number of fsync operations in the bookie journal, the timeout is rounded to the next millisecond.
If the group commit is set to 1ms, that would mean that the effective wait time will be closer to 2ms.
Diffs
-----
bookkeeper-server/src/main/java/org/apache/bookkeeper/bookie/Journal.java 7384330
Diff: https://reviews.apache.org/r/33094/diff/
Testing
-------
Thanks,
Matteo Merli
Re: Review Request 33094: Use nanoseconds to calculate poll timeout
when doing group commit
Posted by Matteo Merli <mm...@yahoo-inc.com>.
> On April 14, 2015, 7:21 a.m., Sijie Guo wrote:
> > bookkeeper-server/src/main/java/org/apache/bookkeeper/bookie/Journal.java, line 824
> > <https://reviews.apache.org/r/33094/diff/1/?file=923861#file923861line824>
> >
> > why change to #get(0)? isn't getFirst clear than #get(0)?
I think this was because this patch was based on some other patch that was using an ArrayList instead of the LinkedList for toFlush. ArrayList and List have no getFirst(), only get(int). I'll change it back to getFirst() for now.
- Matteo
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/33094/#review79996
-----------------------------------------------------------
On April 10, 2015, 9:45 p.m., Matteo Merli wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/33094/
> -----------------------------------------------------------
>
> (Updated April 10, 2015, 9:45 p.m.)
>
>
> Review request for bookkeeper.
>
>
> Bugs: BOOKKEEPER-850
> https://issues.apache.org/jira/browse/BOOKKEEPER-850
>
>
> Repository: bookkeeper-git
>
>
> Description
> -------
>
> When using group commit to reduce the number of fsync operations in the bookie journal, the timeout is rounded to the next millisecond.
> If the group commit is set to 1ms, that would mean that the effective wait time will be closer to 2ms.
>
>
> Diffs
> -----
>
> bookkeeper-server/src/main/java/org/apache/bookkeeper/bookie/Journal.java 7384330
>
> Diff: https://reviews.apache.org/r/33094/diff/
>
>
> Testing
> -------
>
>
> Thanks,
>
> Matteo Merli
>
>
Re: Review Request 33094: Use nanoseconds to calculate poll timeout
when doing group commit
Posted by Sijie Guo <gu...@gmail.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/33094/#review79996
-----------------------------------------------------------
bookkeeper-server/src/main/java/org/apache/bookkeeper/bookie/Journal.java
<https://reviews.apache.org/r/33094/#comment129729>
why change to #get(0)? isn't getFirst clear than #get(0)?
bookkeeper-server/src/main/java/org/apache/bookkeeper/bookie/Journal.java
<https://reviews.apache.org/r/33094/#comment129730>
why #get(0)
- Sijie Guo
On April 10, 2015, 9:45 p.m., Matteo Merli wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/33094/
> -----------------------------------------------------------
>
> (Updated April 10, 2015, 9:45 p.m.)
>
>
> Review request for bookkeeper.
>
>
> Bugs: BOOKKEEPER-850
> https://issues.apache.org/jira/browse/BOOKKEEPER-850
>
>
> Repository: bookkeeper-git
>
>
> Description
> -------
>
> When using group commit to reduce the number of fsync operations in the bookie journal, the timeout is rounded to the next millisecond.
> If the group commit is set to 1ms, that would mean that the effective wait time will be closer to 2ms.
>
>
> Diffs
> -----
>
> bookkeeper-server/src/main/java/org/apache/bookkeeper/bookie/Journal.java 7384330
>
> Diff: https://reviews.apache.org/r/33094/diff/
>
>
> Testing
> -------
>
>
> Thanks,
>
> Matteo Merli
>
>
Re: Review Request 33094: Use nanoseconds to calculate poll timeout
when doing group commit
Posted by Sijie Guo <gu...@gmail.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/33094/#review80920
-----------------------------------------------------------
Ship it!
Ship It!
- Sijie Guo
On April 14, 2015, 5:20 p.m., Matteo Merli wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/33094/
> -----------------------------------------------------------
>
> (Updated April 14, 2015, 5:20 p.m.)
>
>
> Review request for bookkeeper.
>
>
> Bugs: BOOKKEEPER-850
> https://issues.apache.org/jira/browse/BOOKKEEPER-850
>
>
> Repository: bookkeeper-git
>
>
> Description
> -------
>
> When using group commit to reduce the number of fsync operations in the bookie journal, the timeout is rounded to the next millisecond.
> If the group commit is set to 1ms, that would mean that the effective wait time will be closer to 2ms.
>
>
> Diffs
> -----
>
> bookkeeper-server/src/main/java/org/apache/bookkeeper/bookie/Journal.java 7384330
>
> Diff: https://reviews.apache.org/r/33094/diff/
>
>
> Testing
> -------
>
>
> Thanks,
>
> Matteo Merli
>
>
Re: Review Request 33094: Use nanoseconds to calculate poll timeout
when doing group commit
Posted by Matteo Merli <mm...@yahoo-inc.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/33094/
-----------------------------------------------------------
(Updated April 14, 2015, 5:20 p.m.)
Review request for bookkeeper.
Changes
-------
changed back to use getFirst()
Bugs: BOOKKEEPER-850
https://issues.apache.org/jira/browse/BOOKKEEPER-850
Repository: bookkeeper-git
Description
-------
When using group commit to reduce the number of fsync operations in the bookie journal, the timeout is rounded to the next millisecond.
If the group commit is set to 1ms, that would mean that the effective wait time will be closer to 2ms.
Diffs (updated)
-----
bookkeeper-server/src/main/java/org/apache/bookkeeper/bookie/Journal.java 7384330
Diff: https://reviews.apache.org/r/33094/diff/
Testing
-------
Thanks,
Matteo Merli