Currently in PTL tests we do not have microsecond logging turned while during running tests. This causes difficulty to differentiate the message log times between PTL and PBS log files.
We are proposing to turn microsecond logging on by default in PTL. This will help us compare between PTL execution time and PBS event time better.
Please find the design for review here - https://pbspro.atlassian.net/wiki/spaces/PD/pages/1322319874/Enabling+microsecond+logging+by+default+for+PTL
Thanks for proposing this @tejas.pai, this will be great for PTL testing.
Just 1 comment:
I don’t think you need changes to interfaces 3, 4, and 5, the changes to interfaces 1 & 2 should take care of setting PBS_LOG_HIGHRES_TIMESTAMP when we do a revert.
we would need to make make changes to interfaces 3, 4, 5. These interfaces are used in called in revert_pbsconf() which in turn is called in setup(). If we do not update them for microsecond logging then any changes in the test to “PBS_LOG_HIGHRES_TIMESTAMP” will remain when TestSelf.setUp() is called
Proposal looks good Tejas
But your are changing revert_pbsconf() to add ’ PBS_LOG_HIGHRES_TIMESTAMP’ to the list of values to set, and _get_dflt_pbsconfval() to return “1” for PBS_LOG_HIGHRES_TIMESTAMP. So, interfaces 3, 4 & 5 should just use that list and set PBS_LOG_HIGHRES_TIMESTAMP correctly right?
Ok no, i saw the code again, and you are right, you do need to change it to handle the case where a test sets it to 0. Sorry, my bad