On Thu, 2 Nov 2006, Rich Morgan wrote:
>> I run a disk-to-disk copy every morning at 5am, invoked by cron. Every
>> now and then, I see a drop in throughput (today it was from ~45 MB/s to
>> ~34 MB/s, from 5:01:30 to 5:08:30). How can I find out what's running at
>> this time and slowing down the copy? I already found the stuff in
>> /etc/crontab, and I moved that from 6:xx to 7:xx. I checked crontabs for
>> root and eben (news has none), nothing about that time.
>
> Deja vu. Wasn't this topic discussed earlier?
Sorta. I had varying overall throughput, but didn't keep a log so I didn't
know at what time during the copy it happened.
> Anyway, you could set a cron job to do a ps aux > pre_dd_process_list.txt
> before you do your dd to capture a snapshot of what's running and the CPU
> utilization... just a thought.
That's my proposed "brute force" method. Since I'm kicking dd every 30s to
get the throughput, I can run ps at the same time. Any way of associating
a given running process with (number of) outstanding I/O events?
Maybe something in /proc?
--
-eben QebWenE01R@vTerYizUonI.nOetP royalty.no-ip.org:81
Your pretended fear lest error might step in is like the man who
would keep all wine out of the country lest men should be drunk.
-- Oliver Cromwell
-----------------------------------------------------------------------
This list is provided as an unmoderated internet service by Networked
Knowledge Systems (NKS). Views and opinions expressed in messages
posted are those of the author and do not necessarily reflect the
official policy or position of NKS or any of its employees.
This archive was generated by hypermail 2.1.3 : Fri Aug 01 2014 - 18:59:07 EDT