Hide Forgot
Description of problem: oprofiled did not start on Nehalem-EX platform. Then profiler testing FAIL. Test Parameters: DEBUG=off RUNMODE=normal OUTPUTFILE=/var/log/v7/runs/1/profiler/output.log DEVICE= TESTSERVER=unknown using linux image /usr/lib/debug/lib/modules/2.6.32-114.0.1.el6.x86_64/vmlinux Using Linux image /usr/lib/debug/lib/modules/2.6.32-114.0.1.el6.x86_64/vmlinux Subtest Reset: ==== START: Errors during reset may be ignored. ==== Warning: "opcontrol --shutdown" has output on stderr Verified data has beed removed ^^^^ END: Errors during reset may be ignored. ^^^^ PASS Subtest Start Daemon: starting opcontrold Using default event: CPU_CLK_UNHALTED:100000:0:1:1 Error: counter 0 not available nmi_watchdog using this resource ? Try: opcontrol --deinit echo 0 > /proc/sys/kernel/nmi_watchdog Error: oprofiled did not start FAIL Version-Release number of selected component (if applicable): RHEL6.1-20110210.1-x86_64 v7-1.3-10 How reproducible: always Steps to Reproduce: 1. provision RHEL6.1-20110210.1-x86_64 on Nehalem-EX platform 2. install v7 and dependencies 3. run profiler testing Actual results: FAIL Expected results: PASS Additional info:
# vim /etc/modprobe.d/modprobe.conf options oprofile timer=1 Then profiler testing PASS.
(In reply to comment #2) > # vim /etc/modprobe.d/modprobe.conf > options oprofile timer=1 > > Then profiler testing PASS. What was the state of the system for the failure? did the file /etc/modprobe.d/modprobe.conf exist? what did it contain? Thanks!
The file did not exist before. I create the file according to Comment #1, and it work.
*** This bug has been marked as a duplicate of bug 683176 ***
Reopening. The cause is bz 683176, however this is the v7 side of that bug. v7 needs to disabled the nmi_watchdog before running the oprofile service, then run the test portion and upon exit restore the nmi_watchdog state. This will avoid the resource conflict introduced by that change.
Created attachment 483893 [details] revised profiler test, resetting NMI watchdog This revision also changes test flow and logic with respect to forced timer configuration.
Test runs of the above patch still fail on the RHEL 6.1 system I'm running. The "report" subtest has opreport producting an error: "opreport" has output on stderr Overflow stats not available
Will, is it expected behavior that when nmi_watchdog is disabled to release the required timer that opreport would output "Overflow stats not available"? Greg, is there any regular output or just this stderr message?
Looking through the code this appears to be produced when there isn't a /var/lib/oprofile/samples/current/stats/ directory. This output is just going to stderr. This has been removed in later versions of oprofile: http://oprofile.git.sourceforge.net/git/gitweb.cgi?p=oprofile/oprofile;a=commit;h=3cb5ede4de23f32ae57f2f7f50a5642edc33faa6 Looks like you could ignore this message.
Created attachment 485599 [details] revised profiler test, resetting NMI watchdog, ignoring spurious opreport stderr output.
Technical note added. If any revisions are required, please edit the "Technical Notes" field accordingly. All revisions will be proofread by the Engineering Content Services team. New Contents: In v7 1.2, oprofiled could not started on some systems, this issue has been fixed in v7 1.3, now oprofiled can be started on these systems.
An advisory has been issued which should help the problem described in this bug report. This report is therefore being closed with a resolution of ERRATA. For more information on therefore solution and/or where to find the updated files, please follow the link below. You may reopen this bug report if the solution does not work for you. http://rhn.redhat.com/errata/RHBA-2011-0497.html