FYI. I created a test to reproduce the behavior.
https://github.com/apache/bookkeeper/pull/677
- Sijie
On Tue, Oct 24, 2017 at 3:41 PM, Sijie Guo wrote:
> I filed an issue for the problem discussed here: https://github.com/
> apache/bookkeeper/issues/659
>
> - Sijie
>
> On Tue, Oct 24, 2017 at
On Oct 25, 2017 12:53 AM, "Enrico Olivelli - Diennea" <
enrico.olive...@diennea.com> wrote:
Il giorno mar, 24/10/2017 alle 15.41 -0700, Sijie Guo ha scritto:
I filed an issue for the problem discussed here:
https://github.com/apache/bookkeeper/issues/659
Sijie, Charan, thank you for catching th
Il giorno mar, 24/10/2017 alle 15.41 -0700, Sijie Guo ha scritto:
I filed an issue for the problem discussed here:
https://github.com/apache/bookkeeper/issues/659
Sijie, Charan, thank you for catching this
IMHO we should take options 1 o 2 as Sijie proposed
Sijie,
Do you have an approximate ide
I filed an issue for the problem discussed here:
https://github.com/apache/bookkeeper/issues/659
- Sijie
On Tue, Oct 24, 2017 at 2:06 PM, Charan Reddy G
wrote:
> Sure Sijie and Ivan. It sounds appropriate.
>
> Thanks,
> Charan
>
> On Tue, Oct 24, 2017 at 1:00 AM, Sijie Guo wrote:
>
>> Yes, we
Sure Sijie and Ivan. It sounds appropriate.
Thanks,
Charan
On Tue, Oct 24, 2017 at 1:00 AM, Sijie Guo wrote:
> Yes, we should fix this ASAP.
>
> Charan, what is your opinion on the fix?
>
> - Sijie
>
> On Tue, Oct 24, 2017 at 12:12 AM, Ivan Kelly wrote:
>
>> > Use the approach that I used befo
Yes, we should fix this ASAP.
Charan, what is your opinion on the fix?
- Sijie
On Tue, Oct 24, 2017 at 12:12 AM, Ivan Kelly wrote:
> > Use the approach that I used before (as in the old commits at twitter’s
> > branch). Ledger storage is responsible for instantiating the checkpoints.
> I would
> Use the approach that I used before (as in the old commits at twitter’s
> branch). Ledger storage is responsible for instantiating the checkpoints.
I would go with this approach. It was probably me that asked for the
changes in the other direction before, but I can't remember why I
asked for them
Hi all,
Sorry for late response. I apologize for being late. I have spent my time
over the weekend going through all the history about the checkpoint
changes, because the implementation in current master doesn’t match what I
had in my mind. Now I think I have all the details ready for explain.
Th
On Tue, Oct 17, 2017 at 11:46 AM, Charan Reddy G
wrote:
> @Sijie..Did you get chance to go through the scenario/code path.
>
Just come back to US today. I will check and get back to you.
>
> @JV..will create a bug, once I get clarity.
>
> Thanks,
> Charan
>
> On Fri, Oct 13, 2017 at 5:19 PM, V
@Sijie..Did you get chance to go through the scenario/code path.
@JV..will create a bug, once I get clarity.
Thanks,
Charan
On Fri, Oct 13, 2017 at 5:19 PM, Venkateswara Rao Jujjuri wrote:
> Charan this looks like an issue to me. Do we have a defect/issue opened?
>
> On Fri, Oct 13, 2017 at 4:
Charan this looks like an issue to me. Do we have a defect/issue opened?
On Fri, Oct 13, 2017 at 4:18 PM, Sijie Guo wrote:
> Charan,
>
> Didn't mean to say the logic is correct. I was just trying to point out
> the points that I remembered for checkpoint.
>
> I am currently traveling, so I don't
Charan,
Didn't mean to say the logic is correct. I was just trying to point out the
points that I remembered for checkpoint.
I am currently traveling, so I don't have code available to check the
sequence. I can check the logic when I am close to the laptop.
Sijie
On Oct 14, 2017 6:11 AM, "Char
Hey Sijie,
I'm not questioning the semantics of checkpoint or the optimization which
was added with Bookkeeper-564. But my concern is are we sure, checkpoint
logic/code is correct and "marker is only updated when all the entries
added before are persisted.", in the case of SortedLedgerStorage. Can
The core of the checkpoint is:
- marker is only updated when all the entries added before are persisted.
That means it doesn't affect correctness if entries added after are flushed.
- the flush in entry log files is just writing data to filesystem. The real
fsync happens after checkpoint. The sep
14 matches
Mail list logo
