If you want to ask that question on this list you need to be starting 8500 guests on one VM each
running some Linux tasks. We can help you understand that performance problem. If you have one. :)
As you are running 8500 tasks under one guest you might get better help on the LINUX-390 list. Your
problem is likely to be something internal in the Linux guest and not a VM problem.
[EMAIL PROTECTED] wrote:
This may not be the correct place to ask this question, but Google isn't
helping a ton and I can't bear the thought of posting into most other Linux
fora, so I'm going to run this past you folks first.
To cut to the chase, as part of an effort to figur out why we had trouble with
a prof-of-concept application that ate every CPU cycle our IFL could give it, I
have written a couple of small Ruby scripts as a stress-testing mechanism. The
basic idea is:
* Create a small test database on our problematic MySQL image. (Database =
100,000 rows, each of which has a numeric key and one field consisting of 30
random alphabetic characters.) this part is fine.
* On another guest, fork a bunch of processes, each of which will read a random
row from the database, generate another random 30-character string, and update
the record.
This procedure goes fine as long as I fork a few thousand processes. Once I
reach 8500 or so, however, I start receiving this:
Resource temporarily unavailable - fork(2) (Errno::EAGAIN)
According to everything I can find, EAGAIN on fork(2) indicates that the system can not allocate sufficient memory to create the child process, but if I issue "free -m" while my stress test script is running I show plenty of available memory.
Am I hitting a per-user process limit or some such? Any ideas?
TIA,
Jon
--
Stephen Frazier
Information Technology Unit
Oklahoma Department of Corrections
3400 Martin Luther King
Oklahoma City, Ok, 73111-4298
Tel.: (405) 425-2549
Fax: (405) 425-2554
Pager: (405) 690-1828
email: stevef%doc.state.ok.us
