Hi folks,

Now we've gone to 14.03 and have reliable memory reporting via cgroups
I'm wondering what the best way is to get a single, overall (high water
mark) memory usage value for a job for our accounting database.

Using the -X option to sacct to summarise usage for job steps into the
overall job doesn't seem to report any memory used, so it appears we'll
need to parse every job step.

I'm thinking of a simple workflow of the form:

* If there's only a .batch step then use that memory usage
* If there's only a .batch step and a .0 step then add them up
* If there's more then pick the high water mark for all the
  steps and add that to the batch job.

Does that sound about right?

It doesn't need to be byte-accurate, this is more to give users a good
indication of how much memory their jobs are using.

All the best,
Chris
-- 
 Christopher Samuel        Senior Systems Administrator
 VLSCI - Victorian Life Sciences Computation Initiative
 Email: [email protected] Phone: +61 (0)3 903 55545
 http://www.vlsci.org.au/      http://twitter.com/vlsci

Reply via email to