Hi,
Let me first off say I hugely appreciate Ganglia. In the old spirit of
reporting glitches so others may benefit, I'd like to add this brief update
to my earlier notes on installing on CentOS. This time it's a
build-from-source install on Ubuntu 10.04 LTS server.
This is based on "./configure --sysconfdir=/etc/ganglia".
1. Libraries to install before building:
libapr1-dev, libconfuse-dev, libpcre3-dev, and most probably librrd-dev
(there was already a bunch of rrd stuff on my reference system, so not
postive the last is the precise .deb required)
2. Manual steps to take after "make install," due to incompleteness there
(same as with CentOS 5.5):
cp mans/* /usr/share/man/man1
cp gmond/gmond.conf.5 /usr/share/man/man5
cp gmond/gmond.init /etc/init.d/gmond
mkdir /etc/ganglia
cd gmond/modules
cp -a conf.d /etc/ganglia
rm /etc/ganglia/conf.d/example.conf
cd /etc/ganglia
gmond -t > gmond.conf
3. If you'd like to see gmond segfault, after the "cp -a conf.d etc/ganglia"
/step, but before the "gmond -t > gmond.conf" step, issue a
"gmond -m". Now, since "gmond -t merely makes explicit the implicit gmond
configuration settings, why should it segfault just if the module
conf.d files are there when the gmond.conf isn't yet??
4. Create an Ubuntu-style /etc/init.d/gmond file. This is left as something
of an exercise for the reader. However, at then end here you'll find a
stripped-down version that works for me - but beware it's inelegant and
not rigorously tested.
5. Haven't built the --with-gmetad version for the master. I'd expect that
will, as on CentOS, manage to create the /etc/ganglia directory and
perhaps put the gmetad.conf file there - in which case the "mkdir
/etc/ganglia" step in (2) here will be unnecessary.
Question for the Ganglia maintainers: Is it by design or oversight that
"make install" doesn't get most of the .conf files set up in sysconfdir?
When I posted my prior notes to the CentOS list I got seriously flamed for
building from tar rather than running through rpmbuild. I hope building from
tar isn't effectively depricated, since in my experience it's always been
the best way to be sure of getting the most recent version of a critical
daemon right. My rule of thumb is to stay with distros for anything they've
done well and are current enough with, and build-by-hand anything where
their compile options or less-than-current version doesn't fit local needs.
That's generally at most a few daemons on any particular system. I always
leave the libraries and so on stock distro. The storm from some of my peers,
who now regard this as an "inproper" method, surprised me.
So does my experience count as finding a bug which should be reported in the
3.1.7 makefile (not to mention that segfault), or is building from tar
something the Ganglia maintainers feel should be discouraged? For most major
projects it remains the most-supported installation path, as it has long
been.
Best regards,
Whit
=== draft, skeletal Ubuntu-style /etc/init.d/gmond file, use at own risk ===
#!/bin/sh
#
# description: gmond startup script
#
GMOND=/usr/sbin/gmond
. /lib/lsb/init-functions
case "$1" in
start)
log_daemon_msg "Starting GANGLIA ... " "gmond"
if start-stop-daemon --start --oknodo --quiet --exec ${GMOND}; then
log_end_msg 0
else
log_end_msg 1
fi
;;
stop)
log_daemon_msg "Stopping GANGLIA ... " "gmond"
if start-stop-daemon --stop --oknodo --quiet -n gmond; then
log_end_msg 0
else
log_end_msg 1
fi
;;
restart|reload)
$0 stop
$0 start
;;
*)
echo "Usage: $0 {start|stop|restart}"
exit 1
esac
exit 0
------------------------------------------------------------------------------
ThinkGeek and WIRED's GeekDad team up for the Ultimate
GeekDad Father's Day Giveaway. ONE MASSIVE PRIZE to the
lucky parental unit. See the prize list and enter to win:
http://p.sf.net/sfu/thinkgeek-promo
_______________________________________________
Ganglia-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ganglia-general
