Hooking into failed slave launches?

classic Classic list List threaded Threaded
5 messages Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate
star

Hooking into failed slave launches?

Tom Rini
Hey all,

I'm working on extending ComputerListener / SlaveComputer so that
plugins can implement additional policy and veto bringing up slaves that
are otherwise in demand.  The modifications to the core are:
https://github.com/trini/jenkins/commit/28a0a92be9c6f6a55c6cd22d1d19575d2f98bb39

And I've got that working locally with some modifications to my lab
manager plugin locally to limit the number of slaves online at once.
The problem I'm having now is that in addition to needing to change my
plugin to just keep waiting for the slave to come up after we send the
request, it'd be good if I could reliably hook into and find out when
launching the slave agent failed.  Is there a place I can do that
already today?  Thanks!

--
Tom Rini
Mentor Graphics Corporation
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate
star

Re: Hooking into failed slave launches?

kohsuke Kawaguchi (CB)
On 03/07/2011 10:34 AM, Tom Rini wrote:

> Hey all,
>
> I'm working on extending ComputerListener / SlaveComputer so that
> plugins can implement additional policy and veto bringing up slaves that
> are otherwise in demand.  The modifications to the core are:
> https://github.com/trini/jenkins/commit/28a0a92be9c6f6a55c6cd22d1d19575d2f98bb39
>
> And I've got that working locally with some modifications to my lab
> manager plugin locally to limit the number of slaves online at once.
> The problem I'm having now is that in addition to needing to change my
> plugin to just keep waiting for the slave to come up after we send the
> request, it'd be good if I could reliably hook into and find out when
> launching the slave agent failed.  Is there a place I can do that
> already today?  Thanks!

Perhaps SlaveComputer._connect(boolean) can be slightly modified to
report the activity and its outcome?



--
Kohsuke Kawaguchi | CloudBees, Inc. | http://cloudbees.com/
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate
star

Re: Hooking into failed slave launches?

Tom Rini
On 03/08/2011 02:59 PM, Kohsuke Kawaguchi wrote:

> On 03/07/2011 10:34 AM, Tom Rini wrote:
>> Hey all,
>>
>> I'm working on extending ComputerListener / SlaveComputer so that
>> plugins can implement additional policy and veto bringing up slaves that
>> are otherwise in demand. The modifications to the core are:
>> https://github.com/trini/jenkins/commit/28a0a92be9c6f6a55c6cd22d1d19575d2f98bb39
>>
>>
>> And I've got that working locally with some modifications to my lab
>> manager plugin locally to limit the number of slaves online at once.
>> The problem I'm having now is that in addition to needing to change my
>> plugin to just keep waiting for the slave to come up after we send the
>> request, it'd be good if I could reliably hook into and find out when
>> launching the slave agent failed. Is there a place I can do that
>> already today? Thanks!
>
> Perhaps SlaveComputer._connect(boolean) can be slightly modified to
> report the activity and its outcome?

OK.  Would you like that as a separate commit to the above or in the
same?  And is the above otherwise what you're looking for in a pull
request in terms of code comments / commit message?

--
Tom Rini
Mentor Graphics Corporation
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate
star

Re: Hooking into failed slave launches?

kohsuke Kawaguchi (CB)

I've made the additional change and pushed it. This change will be in 1.402.

On 03/08/2011 02:46 PM, Tom Rini wrote:

> On 03/08/2011 02:59 PM, Kohsuke Kawaguchi wrote:
>>  On 03/07/2011 10:34 AM, Tom Rini wrote:
>>>  Hey all,
>>>
>>>  I'm working on extending ComputerListener / SlaveComputer so that
>>>  plugins can implement additional policy and veto bringing up slaves that
>>>  are otherwise in demand. The modifications to the core are:
>>>  https://github.com/trini/jenkins/commit/28a0a92be9c6f6a55c6cd22d1d19575d2f98bb39
>>>
>>>
>>>  And I've got that working locally with some modifications to my lab
>>>  manager plugin locally to limit the number of slaves online at once.
>>>  The problem I'm having now is that in addition to needing to change my
>>>  plugin to just keep waiting for the slave to come up after we send the
>>>  request, it'd be good if I could reliably hook into and find out when
>>>  launching the slave agent failed. Is there a place I can do that
>>>  already today? Thanks!
>>
>>  Perhaps SlaveComputer._connect(boolean) can be slightly modified to
>>  report the activity and its outcome?
>
> OK.  Would you like that as a separate commit to the above or in the
> same?  And is the above otherwise what you're looking for in a pull
> request in terms of code comments / commit message?
>


--
Kohsuke Kawaguchi | CloudBees, Inc. | http://cloudbees.com/
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate
star

Re: Hooking into failed slave launches?

Tom Rini
On 03/09/2011 02:59 PM, Kohsuke Kawaguchi wrote:
>
> I've made the additional change and pushed it. This change will be in
> 1.402.
>

Great, thanks alot!

> On 03/08/2011 02:46 PM, Tom Rini wrote:
>> On 03/08/2011 02:59 PM, Kohsuke Kawaguchi wrote:
>>> On 03/07/2011 10:34 AM, Tom Rini wrote:
>>>> Hey all,
>>>>
>>>> I'm working on extending ComputerListener / SlaveComputer so that
>>>> plugins can implement additional policy and veto bringing up slaves
>>>> that
>>>> are otherwise in demand. The modifications to the core are:
>>>> https://github.com/trini/jenkins/commit/28a0a92be9c6f6a55c6cd22d1d19575d2f98bb39
>>>>
>>>>
>>>>
>>>> And I've got that working locally with some modifications to my lab
>>>> manager plugin locally to limit the number of slaves online at once.
>>>> The problem I'm having now is that in addition to needing to change my
>>>> plugin to just keep waiting for the slave to come up after we send the
>>>> request, it'd be good if I could reliably hook into and find out when
>>>> launching the slave agent failed. Is there a place I can do that
>>>> already today? Thanks!
>>>
>>> Perhaps SlaveComputer._connect(boolean) can be slightly modified to
>>> report the activity and its outcome?
>>
>> OK. Would you like that as a separate commit to the above or in the
>> same? And is the above otherwise what you're looking for in a pull
>> request in terms of code comments / commit message?
>>
>
>


--
Tom Rini
Mentor Graphics Corporation
Loading...