It should work to set up a subversion repository - perhaps a separate
one for each developer - where they can commit the binaries and
jenkins will notice and queue the job. It is difficult to ever
remove anything from a subversion repository, so you'd want to be able
to delete the whole thing periodically and start over. If the changes
between runs are small, I wouldn't zip the files - just let subversion
manage the binary diffs between revisions.
--
Les Mikesell
lesmi...@gmail.com
Because your test run is time-consuming and you probably have multiple
developers, you should allow Jenkins to run multiple builds of this
job in parallel. You get this by checking the box "Execute concurrent
builds if necessary". Also tune the number of executors according to
the number of tests you think your server can run in parallel.
You might find that you need to run more parallel test runs than your
server can handle. Then you can get another server and configure
Jenkins to run a slave on the new server and thus increase the
capacity of your Jenkins.
As you are probably aware already, your most pressing problem is the
long test duration. When the tests take so long, the developers invent
new ways to avoid running them and the ones who run them waste a full
day waiting for feedback.
If I were you, immediately after setting up the test system described
above, I would start working on being able to run the tests in
parallel.
Where I work every developer has their own personal Jenkins job. The
Jenkins job is set to build from the developer's personal git repo and
run tests. (You could maybe use work-in-progress (wip) branches if you
use Subversion.) The test run deploys a cluster of vms, installs the
build artifacts on the vms and runs tests as much in parallel as
possible. This takes about 30 minutes.
A developer is only allowed to submit his changes to the main
repository/master/trunk (take your pick) if he can prove his changes
do not break anything by showing a build of his personal Jenkins job
that has tested his changes.
-- Sami
2012/3/11 AndyNic <andy.nich...@gmail.com>: