https://bugs.koha-community.org/bugzilla3/show_bug.cgi?id=18055 Jonathan Druart <jonathan.druart@bugs.koha-community.org> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |jonathan.druart@bugs.koha-c | |ommunity.org --- Comment #26 from Mason James <mtj@kohaaloha.com> --- Created attachment 64322 --> https://bugs.koha-community.org/bugzilla3/attachment.cgi?id=64322&action=edit Bug 18055 - Speed up '00-strict.t' test, with Parallel::ForkManager add KOHA_PROVE_CPUS envar to manually set number of cpus to test.. 1/ set number of cpus to 1 $ export KOHA_PROVE_CPUS=1 2/ observe 1 cpus are used $ time prove -v t/db_dependent/00-strict.t t/db_dependent/00-strict.t .. Using 1 CPUs... 3/ unset number of cpus $ export KOHA_PROVE_CPUS= 4/ observe all available cpus are used $ time prove -v t/db_dependent/00-strict.t t/db_dependent/00-strict.t .. Using 32 CPUs... --- Comment #27 from Mason James <mtj@kohaaloha.com> --- (In reply to Mason James from comment #21)
(In reply to Olli-Antti Kivilahti from comment #19)
Hi!
Pretty good improvement.
What concerns me is how well does this module detect the available CPU cores?
For example on LXC all the cores are visible to the test runner.
There can be resource limits, but we might get a situation where we fork 32 test runners because the LXC container is ran on a server with 32 cores. However resource limits prevent utilization of more than 4 on this LXC-container.
This is not good.
Olli, We can add an enviroment variable to manually limit the number of CPUS, to handle your situation
I can add a patch for that, no problem
I've added this patch --- Comment #28 from Jonathan Druart <jonathan.druart@bugs.koha-community.org> --- Could we use the same method for t/00-testcritic.t, t/00-merge-conflict-markers.t, xt/author/valid-templates.t and xt/author/translatable-templates.t as well? Any volunteers? -- You are receiving this mail because: You are watching all bug changes.