[ale] Tool search

[Ed Cashin]

I was watching for responses.  This doesn't sound simple at all to me.

For a while they were adding dtrace to Linux, and I think Oracle Linux had
it.  I don't know whether it made it into the mainline, but if so, you
could probably put together what you need using dtrace.  You'd have to
consider all the ways (direct IO, minor page faults) bytes can get into
RAM, though, and filter out stuff you're not interested in (local disk,
etc.).

If there's no special tool and no dtrace, ftrace might be of interest to
you.


On Thu, Oct 15, 2020 at 5:51 PM Jim Kinney via Ale <ale at ale.org> wrote:

> I looked at bonnie++ but couldn't figure a way have it basically monitor
> whatever is being opened.
>
> This is an hpc stack with some users opening multiple TB in tens of
> thousands files for a single job.
>
> I'm still pretty convinced the solution involves pulling data from a
> cgroup structure. I have some cgroup contraints on memory per user. Adding
> cores per user next since some people don't spec how many cores they want
> to schedule and their code eventually says "all available". So no specific
> request means 1 core.
>
> On October 15, 2020 3:52:48 PM EDT, Derek Atkins <derek at ihtfp.com> wrote:
>>
>> Can you run something like bonnie++ as the user-in-question?
>> -derek
>>
>> On Thu, October 15, 2020 3:42 pm, Jim Kinney via Ale wrote:
>>
>>>  I have need to measure NFS mount bandwidth on a per user basis. Nfsiostat
>>>  is aggregate over a mount point. Users may interact with multiple mount
>>>  points for any task. Multiple may be on the same mount point at the same
>>>  time.
>>>
>>>  Looking at cgroups as a per user isolation method. Not seeing how to get
>>>  "bytes moved into RAM. Probably stupid simple once I see it...
>>>  --
>>>  Computers amplify human error
>>>  Super computers are really
>>>  cool_______________________________________________
>>>  Ale mailing list
>>>  Ale at ale.org
>>>  https://mail.ale.org/mailman/listinfo/ale
>>>  See JOBS, ANNOUNCE and SCHOOLS lists at
>>>  http://mail.ale.org/mailman/listinfo
>>>
>>>
>>
> --
> Computers amplify human error
> Super computers are really cool
> _______________________________________________
> Ale mailing list
> Ale at ale.org
> https://mail.ale.org/mailman/listinfo/ale
> See JOBS, ANNOUNCE and SCHOOLS lists at
> http://mail.ale.org/mailman/listinfo
>


-- 
  Ed Cashin <ecashin at noserose.net>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.ale.org/pipermail/ale/attachments/20201016/73dedc24/attachment.html>