MINA 2.0.12

[Emmanuel Lécharny <el...@gmail.com>]

Hi guys,

we just have released MINA 2.0.1 (still to be announced) and Radovan
just found a hug bug in the way we deal with closed session. This is
related to DIRMINA-1001 and DIRMINA-1006. All in all, the code that
manage session removal and the check for the spin-bug detection is not
good, and in some corner cases, we might enter in an infinite loop, with
100% CPU. Fixing that issue was not simple, but once done, Radovan found
that the removal of a session from the queue that manage sessions to be
removed was not properly done, leading to the IoProcessor to never been
shut down (we don'te xit from the infinite select() loop, although it
does not eat any CPU).

Thanks to radovan, we have found many fixes that were applied, and we
are half-confident that this patch fixes most of the problems. Although
we need to conduct some in depth review of the code that deal with
session being closed (and there are various use cases here). This is
what I'm going to do this week-end - if it rains enough so that I can't
be dragged out by my dear and active wife-. I would appreciate any help
in this task. I think that has to be documented somewhere on the web
site, at least to keep a track of the way it works, and also for our users.

In the mean time, I think we outght to cut a release because this is a
nasty bug, and I'd like to get this release out on monday, shortening
the delay for a vote to 48h, if I can push the new version at noon today.

wdyt ?

Re: MINA 2.0.12

[Ashish <pa...@gmail.com>]

+1

On Sat, Jan 30, 2016 at 7:32 AM, Jeff Genender <jg...@apache.org> wrote:
> +1
>
> Jeff
>
>> On Jan 30, 2016, at 1:36 AM, Emmanuel Lécharny <el...@gmail.com> wrote:
>>
>> Hi guys,
>>
>> we just have released MINA 2.0.1 (still to be announced) and Radovan
>> just found a hug bug in the way we deal with closed session. This is
>> related to DIRMINA-1001 and DIRMINA-1006. All in all, the code that
>> manage session removal and the check for the spin-bug detection is not
>> good, and in some corner cases, we might enter in an infinite loop, with
>> 100% CPU. Fixing that issue was not simple, but once done, Radovan found
>> that the removal of a session from the queue that manage sessions to be
>> removed was not properly done, leading to the IoProcessor to never been
>> shut down (we don'te xit from the infinite select() loop, although it
>> does not eat any CPU).
>>
>> Thanks to radovan, we have found many fixes that were applied, and we
>> are half-confident that this patch fixes most of the problems. Although
>> we need to conduct some in depth review of the code that deal with
>> session being closed (and there are various use cases here). This is
>> what I'm going to do this week-end - if it rains enough so that I can't
>> be dragged out by my dear and active wife-. I would appreciate any help
>> in this task. I think that has to be documented somewhere on the web
>> site, at least to keep a track of the way it works, and also for our users.
>>
>> In the mean time, I think we outght to cut a release because this is a
>> nasty bug, and I'd like to get this release out on monday, shortening
>> the delay for a vote to 48h, if I can push the new version at noon today.
>>
>> wdyt ?
>>
>



-- 
thanks
ashish

Blog: http://www.ashishpaliwal.com/blog
My Photo Galleries: http://www.pbase.com/ashishpaliwal

Re: MINA 2.0.12

[Jeff Genender <jg...@apache.org>]

+1

Jeff

> On Jan 30, 2016, at 1:36 AM, Emmanuel Lécharny <el...@gmail.com> wrote:
> 
> Hi guys,
> 
> we just have released MINA 2.0.1 (still to be announced) and Radovan
> just found a hug bug in the way we deal with closed session. This is
> related to DIRMINA-1001 and DIRMINA-1006. All in all, the code that
> manage session removal and the check for the spin-bug detection is not
> good, and in some corner cases, we might enter in an infinite loop, with
> 100% CPU. Fixing that issue was not simple, but once done, Radovan found
> that the removal of a session from the queue that manage sessions to be
> removed was not properly done, leading to the IoProcessor to never been
> shut down (we don'te xit from the infinite select() loop, although it
> does not eat any CPU).
> 
> Thanks to radovan, we have found many fixes that were applied, and we
> are half-confident that this patch fixes most of the problems. Although
> we need to conduct some in depth review of the code that deal with
> session being closed (and there are various use cases here). This is
> what I'm going to do this week-end - if it rains enough so that I can't
> be dragged out by my dear and active wife-. I would appreciate any help
> in this task. I think that has to be documented somewhere on the web
> site, at least to keep a track of the way it works, and also for our users.
> 
> In the mean time, I think we outght to cut a release because this is a
> nasty bug, and I'd like to get this release out on monday, shortening
> the delay for a vote to 48h, if I can push the new version at noon today.
> 
> wdyt ?
>