Changes committed to git://github.com/pauljevans/pcp.git dev
qa/654 | 2 +-
qa/655 | 22 +-
qa/655.out | 1655
++++++++++++----------------------------------
src/pmdas/gfs2/ftrace.c | 3 +-
src/pmdas/gfs2/latency.c | 6 +-
src/pmdas/gfs2/pmda.c | 25 +-
6 files changed, 472 insertions(+), 1241 deletions(-)
commit e3023879da3ebdbe0012db6326a3ff9df8ef81dc
Author: Paul Evans <pevans@xxxxxxxxxx>
Date: Thu Aug 28 17:54:19 2014 +0100
pmdagfs2: Fix issue with RHEL 6.6 ftrace not having irq-info option
The current version of ftrace included in the RHEL 6.6 beta does
not include
the irq-info option as part of its trace function output (default
is include
and enabled with other distributions). This results in the
tracepoint output
given by the trace_pipe being different to what is expected and the
PMDA
hanging on collection of the newly added latency metrics (timeout
and killed
by PMCD).
To tackle this we turn off the irq-info option in versions of
ftrace that
have the option and tweak the expected trace_pipe output so that
the same
output will be given by all distributions if they have irq-info option
support or not.
commit 01bcd0cb503401e13f737cf0d1d8f85800e65538
Author: Paul Evans <pevans@xxxxxxxxxx>
Date: Thu Aug 28 17:52:28 2014 +0100
pmdagfs2: Update qa tests to remove erroneous reference to valgrind
Update qa/655 to remove erroneous reference to valgrind in the echo
text.
Also made a few more corrections to qa/654 as well as updating 655.out.
Have also changed the metrics checked in qa/655 to exclude metrics
which may
change from "no values available" to having metric values based on
filesyste
activity during testing.
Took me longer than it should have to figure out why qa was failing only on
RHEL 6.6 and not on the others, managed to track it down to ftrace missing
the irq-info option which is available and enabled by default on all other
distributions (need to check if this is intended).
The attached fix will sorts out the problem whilst ensuring code parity for
all distributions. I plan on coming back to this in the next couple of weeks
to improve the tracepoint fetch code by switching to using the ftrace binary
interface which should be a cleaner and quicker implementation.
With qa/655, the valgrind references managed to slip in from some other qa
improvements work that I am testing at the moment these aren't ready to be
sent yet; Have removed them whilst making some additional improvements to
qa/655.
All of the patches have been tested and have had covscan run on them. As
always please let me know if there are any issues and feedback is welcome.
Cheers,
Paul
|