Changes committed to git://oss.sgi.com/pcp/pcp.git dev
qa/.gitignore | 2
qa/003 | 1
qa/300 | 21 +
qa/300.out | 17 -
qa/300.out.1 | 17 +
qa/300.out.2 | 23 +-
qa/322 | 2
qa/438 | 15 +
qa/438.out | 103 ---------
qa/438.out.1 | 103 +++++++++
qa/438.out.2 | 129 ++++++++++++
qa/group | 1
qa/src/GNUlocaldefs | 3
qa/src/new_mysql.0 |binary
qa/src/new_mysql.index |binary
qa/src/new_mysql.meta |binary
src/libpcp/src/check-statics | 2
src/pmdas/linux/help | 6
src/pmdas/linux/indom.h | 4
src/pmdas/linux/pmda.c | 430 ++++++++++++++++++++++------------------
src/pmdas/linux/proc_net_snmp.c | 386 ++++++++++++++++++++++++++++++-----
src/pmdas/linux/proc_net_snmp.h | 31 ++
src/pmdas/linux/proc_vmstat.c | 318 +++++++++++++++++++----------
src/pmdas/linux/proc_vmstat.h | 2
src/pmdas/linux/root_linux | 10
src/pmdas/mysql/GNUmakefile | 1
src/pmdas/mysql/migrate.conf | 14 +
27 files changed, 1138 insertions(+), 503 deletions(-)
commit 076259bb197cb33f1eb6d501406ee8395f0ad7e4
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date: Mon Jul 22 10:54:34 2013 +1000
Fix loop termination check for ICMP stats, older kernels
In the reworked /proc/net/snmp parsing code, as a result
of the incsumerrors field being optional (older kernels),
we need to be careful when matching up header and values
columns against the expected set. This fixes incorrect
behaviour where the final field in the row was missed, in
the case where there is a missing field earlier on in the
row (as is the case with the ICMP metric values).
commit 93981600797695ad55b81d08e4693ab3cd18049f
Merge: d3f6725 2da251a
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date: Mon Jul 22 08:42:43 2013 +1000
Merge branch 'dev' of git://oss.sgi.com/kenj/pcp into dev
commit d3f6725ba6caf6998d9d6c04cc2ca801bb855cc3
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date: Mon Jul 22 08:42:25 2013 +1000
Rework (and fix) the /proc/net/snmp metric extraction code
As recently discovered by the Red Hat QE folk (running qa/126
on recent kernels) the /proc/net/snmp file has been extended
to add checksum error counters to several message types. For
most cases this was fine and the code still worked, except in
the case of the Icmp values. In that case, the new value was
added in the middle of the existing set and not the end.
This has exposed an assumption in the code where it chose to
ignore the header information and assumed all new values would
be appended. This is no longer feasible, and is tackled here.
We now parse the headers (every other line) and match them up
to the values on the following lines. The new checksum error
metrics are added and reported as not-available on all kernels
predating this feature.
Implementation-wise, these metrics are handled more like the
vmstat metrics now, so there was some minor cleanup resulting
from a closer look at that code (esp. removed a dup'd global).
Finally, we now had the option of also implementing the icmpmsg
values which previously we could not due to ignoring the header
(these values are more dynamic, and not a fixed set of values).
This also has been done, and a new indom is added to handle the
more dynamic nature of these values (see network.icmpmsg.*).
commit 2da251a2989c19bb0dd1692a13f3db2fcfa595f4
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Mon Jul 22 06:44:55 2013 +1000
mysql pmda metric migration change support
pmlogrewrite config and qa for recent metadata change for mysql pmda.
commit 3aeaba8fd6ab144b18696c2425f37f68f6b504f1
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Sun Jul 21 14:44:08 2013 +1000
check-statics - require_credentials wrong symbol type
Should be b (BSS) not d (data). Fails when HAVE_SECURE_SOCKETS
is undefined.
commit 78be64ac0960d1b94039f16a2abd58ed8fec88b4
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date: Sat Jul 20 13:15:32 2013 +1000
Fix issue where errno is passed out as result of a pmdalinux fetch
Dave observed spurious log messages during some test runs recently,
turned out to be a permissions problem on some statfs paths. We
should not be passing errno out here - its platform specific and it
is unhelpful in general, better to use PM_ERR_INST for those paths,
which will not cause logged messages (unless debug enabled which is
just fine).
commit cbbd831b8ba790f68772fa5b9dc35a019c4dc353
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Fri Jul 19 21:00:37 2013 +1000
qa/322 - add -f for rm $seq.out (first time failure)
commit 52d64727fc09713c849558629ae8e8211f5c73cd
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Fri Jul 19 20:58:40 2013 +1000
qa/300 - filter out literal hostname
commit 36b56c08a8d865bbf56606b82ff5c5171daa0cdb
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Fri Jul 19 20:51:51 2013 +1000
qa/300 - make variant
Extra lines like
TIME start pmlogger from pmlogger_check for host bozo-laptop
in $PCP_LOG_DIR/NOTICES
commit 730b0c8d9dcb1760afec452ba33c0e5913e3fe15
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Fri Jul 19 20:21:38 2013 +1000
qa/003 - hinv.map.lvname may have no values on some platforms
commit 23eb31082fc09143b51a23367f7f538888288757
Merge: f320812 df88a72
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Fri Jul 19 19:53:20 2013 +1000
Merge branch 'dev' of git://oss.sgi.com/pcp/pcp into dev
commit f320812d523b86107ea47e7d071a425685e0035a
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Fri Jul 19 19:46:52 2013 +1000
qa/438 - new variant
Commit 4670b16 change pmdumplog in a subtle way that changed the
initial call to pmSetMode() for this test, and produced slightly
different -Dlog output.
Also a couple more badlen-* archives have been added recently.
|