On May 4, 2004, at 08:42, Mike wrote:
My site has many shell scripts that I want preserved
in CVS. My first thought was to do a 'cvs co' on each
server prior to running cfagent, then copy files to
the proper locations as necessary.
I have a cfengine master and several cfengine slaves. My master does
"cvs co" into /var/cfengine/cvsexport/ and my slaves copy: that.
Clients copy: from the slaves.
Now I'm wondering
about using cfservd with cfagent to get the files
directly from the cvshost.
Seems like it would be cool if cfengine knew how to copy out of CVS
repositories directly. It seems like it would help with rolling out
new
versions of things out in a controlled manner. On the other hand,
having the separate checkout directory is simple and is working fine
for me.
I guess I need to do a
'cvs co' from cron with some frequency to keep the
repository for cfservd current, or is there another
way?
My master updates its own cvs workspace every time its cfagent runs. I
also have CVS run cfagent every time a change is checked in. And I use
cfrun to have the slaves update whenever the master updates. I have
cron run cfexecd every half hour. There are quirks here and there, but
overall it's working out very nicely.
My update.conf includes the following. I would love to see any ideas
for improvement!
control:
cvs_dir = ( /var/cfengine/cvsexport )
shellcommands:
master::
"/usr/bin/cvs -d /var/cvsroot -q checkout ."
chdir=${cvs_dir}
"/usr/local/sbin/cfrun -S -T -- --silent -- slave
/dev/null"
copy:
slave::
${cvs_dir}
dest=${cvs_dir}
server=${server}
recurse=inf
trustkey=on
purge=on
backup=false
timestamps=preserve
inform=on
Mike