End of this story
With this patch the problem does not occur anymore
https://github.com/apache/bookkeeper/pull/1293
The patch does not address directly the problem, the root source is still
unknown, this is very bad. But with that change no error is reported
anymore, so actually it is enough to
2018-03-09 8:59 GMT+01:00 Sijie Guo :
> Sent out a PR for the issues that I observed:
>
> https://github.com/apache/bookkeeper/pull/1240
>
Other findings:
- my problem is not related to jdk9, it happens with jdk8 too
- the "tailing reader" is able to make progress and follow
Sent out a PR for the issues that I observed:
https://github.com/apache/bookkeeper/pull/1240
On Thu, Mar 8, 2018 at 10:47 PM, Sijie Guo wrote:
> So the problem here is:
>
> - a corrupted request failed the V3 request decoder, so bookie switched to
> use v2 request decoder.
So the problem here is:
- a corrupted request failed the V3 request decoder, so bookie switched to
use v2 request decoder. Once the switch happen, the bookie will always use
v2 request decoder decoding v3 request. then all your v3 requests will be
failing with unknown op and trigger the bytebuf
On Thu, Mar 8, 2018 at 1:05 PM, Sijie Guo wrote:
>
> On Thu, Mar 8, 2018 at 12:33 PM Andrey Yegorov
> wrote:
>
>> I am looking more at the PendigAddOp and it looks like, in addition to
>> the case that Sijie has fixed, there is another scenario
Il gio 8 mar 2018, 22:22 Enrico Olivelli ha scritto:
>
>
> Il gio 8 mar 2018, 22:08 Sijie Guo ha scritto:
>
>> This is even interesting with v3 protocol. Currently the v3 protocol
>> copies the byte string and it doesn’t even use pooled buffer. So I am
This is even interesting with v3 protocol. Currently the v3 protocol copies
the byte string and it doesn’t even use pooled buffer. So I am not sure if
the issue comes from bookie or an issue at the netty layer. It would be
good to get a repo.
Sijie
On Thu, Mar 8, 2018 at 12:29 PM Enrico Olivelli
Il gio 8 mar 2018, 21:33 Andrey Yegorov ha
scritto:
> I am looking more at the PendigAddOp and it looks like, in addition to the
> case that Sijie has fixed, there is another scenario where recycler can get
> triggered.
>
I think that this is another issue, but your
I am looking more at the PendigAddOp and it looks like, in addition to the
case that Sijie has fixed, there is another scenario where recycler can get
triggered.
I.e.:
first sendWriteRequest() in PendingAddOp.safeRun() fails in PCBC's
writeAndFlush (channel == null or channel.writeAndFlush
Il gio 8 mar 2018, 20:52 Sijie Guo ha scritto:
> On Thu, Mar 8, 2018 at 7:42 AM, Enrico Olivelli
> wrote:
>
>>
>>
>> 2018-03-08 14:50 GMT+01:00 Ivan Kelly :
>>
>>> It just occurred to me that this could be a problem with the recycler.
On Thu, Mar 8, 2018 at 7:42 AM, Enrico Olivelli wrote:
>
>
> 2018-03-08 14:50 GMT+01:00 Ivan Kelly :
>
>> It just occurred to me that this could be a problem with the recycler.
>> If we recycle a buffer too early, but then keep using it, another user
>>
2018-03-08 14:50 GMT+01:00 Ivan Kelly :
> It just occurred to me that this could be a problem with the recycler.
> If we recycle a buffer too early, but then keep using it, another user
> could pick it up, and between them they could corrupt the data that
> would cause it to be
12 matches
Mail list logo