Hi -
Review sought for freshly rebased pcpfans.git fche/pmmgr code. Just
two nontrivial commits there. (I plan to work more in the near future
on accelerating the 666 test case.)
commit 26de1cabca356ca9b73cc5cd0b79da1da4dc1b58
Author: Frank Ch. Eigler <fche@xxxxxxxxxx>
Date: Fri Jul 4 13:05:13 2014 -0400
pmmgr: handle case of pm{ie,logger} daemon that refuses SIGTERM
In rare cases, it appears possible for a signal to a child
pmie/pmlogger process to be lost, when the pmmgr_daemon dtor wants
to kill it. Like pm*_daily.sh, be assertive with a later SIGKILL,
and don't wait forever.
commit 7d2230f517b5ee51a568ac25e4f4871ddedba42a
Author: Frank Ch. Eigler <fche@xxxxxxxxxx>
Date: Wed Apr 9 13:28:14 2014 -0400
pmmgr testing: quicken, avoid some granularity-edge races
After concerns, the time taken by the pmmgr 666 test case are now
reduced to about 6 minutes.
Changes to pmmgr proper involve active avoidance of granular-mode
period boundaries. pmloggers are instructed to shut down one second
before, and new pmloggers are precluded from launching within that
transitional second. This seems to make the resulting archives'
timespans match the intuitive expetations.
|