I was checking on a number of builds that I have just converted over to hudson.
The code base is quite large (3 GB), so it takes quite a while to get the code
from AccuRev. From the hudson main page it looked as though all the builds had
started, but when I actually check the build directories, only one was being
populated. Checking the job logs, the one that was being populated had a log
file that showed content being populated in the directories. One of the other
jobs was at the "populating workspace" message in the log, and the other was at
the "Authenticating" message.
I waited a number of minutes for the first job to finish populating. Shortly
after completion of the first populate, one of the other jobs started populating
the build directory. Once that one completed populating, the last job started
populating the directory.
All the jobs are set up to poll AccuRev at the same time using the same user,
and build the code if changes in the selected stream are found (also the same
across all the jobs).
Is there any reason why the populating of the workspace is done in a serial
fashion? I had these set up as a cron job before converting to hudson, with the
same configuration (same user populating different directories at the same time,
etc) and never had to serialize the builds.
Given the size of the code base, this could have a negative impact on the
performance of our already lengthy builds.