Faisal,
Thanks for your post.
The problem you face is a very common one. When I tried to implement a fix
for this a few year ago using Hudson, I was unable to do so using a purely
Hudson approach. I tried various approaches, such as using build quiet
times (as documented in the above posts). But I still was receiving the
same errors you report.
What I ended up doing was the following:
1. enable scm polling in all projects (but don't schedule any polling)
2. write an OS shell script (e.g. using bash) that uses wget to access the
poll URL of each project in their dependency order
for example, something like:
wget http://myhudsonserver:port/myjob1/poll?token=myToken
wget http://myhudsonserver:port/myjob2/poll?token=myToken
3. create a new hudson job that calls this shell script every few minutes
The above seemed like a horrible hack at the time, but I've grown to like
it as it makes it easy to disable all scm polling for all jobs when needed
(e.g. we do this sometimes for short periods when doing server maintenance,
or when doing releases of our code). Another advantage of this is I've been
able to segregate the projects that must be built serially (e.g. small,
quickly building framework related projects that most other projects depend
on) from the rest of the projects (that can be built in parallel). The
latter type projects really don't need a strict build order, but I still
wrote a triggering script that triggers all of them (so I can stop and
start all their polling easily as well), and another hudson job that is
triggered once the serial jobs run and that triggers the polling of the
parallel jobs (as this ensures the parallel jobs won't start until all the
serial jobs are complete). The above assumes we organize the hudson jobs
into slaves as follows:
serial slave (1 executor thread) runs the following jobs:
shell script for polling serial jobs
serial job 1
serial job 2
.....
serial job n
shell script for polling parallel jobs
parallel slave (several executor threads) runs the following jobs:
parallel job 1
parallel job 2
.....
parallel job n
In the time since I have implemented the above, I wouldn't be surprised if
someone has found (or added) features or a plugin to hudson or jenkins that
does the above. If so, I would love to hear about it and test it out if
time permits.
Brian
On Saturday, January 12, 2013 12:12:28 PM UTC-6, Faisal Faizan wrote:
>
>
> I have exactly the same situation ....
>
