Closed martin-g closed 1 year ago
Most TIMEOUTs happened during the BUILD step (R CMD build
) which is the step that tends to put the most load on the machine. Don't you want to lower BBS_BUILD_NB_CPU
too?
Done! Let's see! I hope the report won't take more than a day now! :-)
I was only suggesting that you lower BBS_BUILD_NB_CPU
, which you didn't in the first iteration of the PR. But I see that you also lowered BBS_CHECK_NB_CPU
again, this time from 20 to 16, when you already lowered it from 25 to 20 in the first iteration of the PR. That's probably excessive, and it's indeed possible that this will put builds at the risk of not finishing in time, hard to know.
The R CMD check
commands that are run during the CHECK step tend to generate less load than the R CMD build
commands that are run during the BUILD step, which is why we use a higher value for BBS_CHECK_NB_CPU
on the build machines (typically 20% or 25% higher).
This is why here I was suggesting the following values:
export BBS_BUILD_NB_CPU=16 # 32 cores are available
export BBS_CHECK_NB_CPU=20 # 32 cores are available
With the old ones the system is too busy and the builds and checks of some packages fail due to timeouts