You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@aurora.apache.org by Isaac Councill <is...@hioscar.com> on 2014/09/10 20:03:55 UTC

one lock to lock them all

Hi,

I just experienced an issue where a job update was failing and leaving a
lock in place. Expected. What I didn't expect was that while that lock was
in place, it was not possible to interact with any job (old or new) due to
the lock. All the jobs I tested were in the same role and prod environment.

Is this behavior intentional?

I'm working with the 0.5.0-incubating release.

Re: one lock to lock them all

Posted by Isaac Councill <is...@hioscar.com>.
Sounds good to me. I need wickman's announcer change anyway :).

Thanks!

On Wed, Sep 10, 2014 at 5:11 PM, Bill Farner <wf...@apache.org> wrote:

> That patch was definitely not in the release.  Isaac, if you do a build
> after this commit, it will be fixed:
>
> commit b0c6f91
> Author: Bill Farner <wf...@apache.org>
> Date:   Wed Aug 6 18:10:58 2014 -0700
>
>     Fix incorrect join type used in LockMapper.xml.
>
>     Bugs closed: AURORA-640
>
>     Reviewed at https://reviews.apache.org/r/24432/
>
> That said, i suggest you cut a build off HEAD.
>
>
> -=Bill
>
>
> On Wed, Sep 10, 2014 at 11:19 AM, David McLaughlin <da...@dmclaughlin.com>
> wrote:
>
> > Hi Isaac,
> >
> > That behaviour is definitely not intentional. This sounds like it was the
> > bug with an incorrect JOIN type in SQL. The ticket tracking that issue is
> > here:
> >
> > https://issues.apache.org/jira/browse/AURORA-640
> >
> > I'm not sure though if the fix made it in in time for that release.
> >
> >
> > - David
> >
> >
> > On Wed, Sep 10, 2014 at 11:03 AM, Isaac Councill <is...@hioscar.com>
> > wrote:
> >
> > > Hi,
> > >
> > > I just experienced an issue where a job update was failing and leaving
> a
> > > lock in place. Expected. What I didn't expect was that while that lock
> > was
> > > in place, it was not possible to interact with any job (old or new) due
> > to
> > > the lock. All the jobs I tested were in the same role and prod
> > environment.
> > >
> > > Is this behavior intentional?
> > >
> > > I'm working with the 0.5.0-incubating release.
> > >
> >
>

Re: one lock to lock them all

Posted by Bill Farner <wf...@apache.org>.
That patch was definitely not in the release.  Isaac, if you do a build
after this commit, it will be fixed:

commit b0c6f91
Author: Bill Farner <wf...@apache.org>
Date:   Wed Aug 6 18:10:58 2014 -0700

    Fix incorrect join type used in LockMapper.xml.

    Bugs closed: AURORA-640

    Reviewed at https://reviews.apache.org/r/24432/

That said, i suggest you cut a build off HEAD.


-=Bill


On Wed, Sep 10, 2014 at 11:19 AM, David McLaughlin <da...@dmclaughlin.com>
wrote:

> Hi Isaac,
>
> That behaviour is definitely not intentional. This sounds like it was the
> bug with an incorrect JOIN type in SQL. The ticket tracking that issue is
> here:
>
> https://issues.apache.org/jira/browse/AURORA-640
>
> I'm not sure though if the fix made it in in time for that release.
>
>
> - David
>
>
> On Wed, Sep 10, 2014 at 11:03 AM, Isaac Councill <is...@hioscar.com>
> wrote:
>
> > Hi,
> >
> > I just experienced an issue where a job update was failing and leaving a
> > lock in place. Expected. What I didn't expect was that while that lock
> was
> > in place, it was not possible to interact with any job (old or new) due
> to
> > the lock. All the jobs I tested were in the same role and prod
> environment.
> >
> > Is this behavior intentional?
> >
> > I'm working with the 0.5.0-incubating release.
> >
>

Re: one lock to lock them all

Posted by David McLaughlin <da...@dmclaughlin.com>.
Hi Isaac,

That behaviour is definitely not intentional. This sounds like it was the
bug with an incorrect JOIN type in SQL. The ticket tracking that issue is
here:

https://issues.apache.org/jira/browse/AURORA-640

I'm not sure though if the fix made it in in time for that release.


- David


On Wed, Sep 10, 2014 at 11:03 AM, Isaac Councill <is...@hioscar.com> wrote:

> Hi,
>
> I just experienced an issue where a job update was failing and leaving a
> lock in place. Expected. What I didn't expect was that while that lock was
> in place, it was not possible to interact with any job (old or new) due to
> the lock. All the jobs I tested were in the same role and prod environment.
>
> Is this behavior intentional?
>
> I'm working with the 0.5.0-incubating release.
>