We are running perf on a 2.6.32-358.56.1.el6.x86_64 on an 8-core, dual socket machine. We have CPU isolation on CPUs 1-31 (all except for 0), and our application is using CPUs 1-31. A few seconds into running perf with --pid set the entire blade crashes and we have to reset it manually. We don't see any logs in /var/log/messages. Any ideas anyone?
Asked
Active
Viewed 65 times
0
-
Looks like a bug in your VM software. Better ask on ServerFault. – Dummy00001 Nov 09 '15 at 09:07
-
Thanks. When you say VM software, what VM are you talking about? We are running on an ATCA blade. – c-is-best Nov 09 '15 at 09:43
-
Also, why is ServerFault a better place? – c-is-best Nov 09 '15 at 09:43
-
"iso on CPUs 1-31" sounded to me like a "VM image given these CPUs." StackOverflow is wrong place, because your question doesn't involve directly software *development*. On ServerFault there are more people with RHEL/etc expertise and how software runs them. – Dummy00001 Nov 09 '15 at 11:26
-
By iso i meant isolation. sorry for the confusion. perf is a developers tool, not specific to any distribution. it's a kernel tool. i simply gave the specifics of the system. – c-is-best Nov 09 '15 at 20:02
-
But `perf` is a system tool provided to you by the OS distribution. And it crashes the whole system. Choices are: (1) ask on ServerFault, since somebody could have seen the problem before and might have a workaround; (2) report the problem to the RH support. Asking on SO would have been OK, if you have modified the kernel or perf, that have caused the crash, and you need the help figuring it out. – Dummy00001 Nov 10 '15 at 08:41