Re: Strange behavioural differences when invoking cfengine through cfrun

[Eric Dorland]

On Wed, 2005-03-09 at 15:20 -0600, rader@ginseng.hep.wisc.edu wrote:
> PLEASE complain LOUDLY to the list about this bug!

Yes, this is a really annoying bug! It was really hard to figure out as
well. I've taken out the SIGCHLD line and now it works just fine with no
ill effects. Isn't SIG_IGN the default anyway? Why not just install a
signal handler that calls wait? 

> Alas I suspect Mark doesn't have any RHEL3 systems and since
> only two of us have complained, he's not very motivated to
> fix it.

I also have a YellowDog 4 system, and it's broken there too. 

> I don't know about you, but not being able to push and
> not being able to debug with "cfrun somehost" really sucks.

It really sucks, especially since I'm rolling out things right now and
doing a lot of debugging...

> - - -
> 
>  > ---- Original Message ----
>  > From: Eric Dorland
>  > Alright, I did a bit more research and I discovered some more puzzling
>  > things. From the debug ouput I have this:
>  > 
>  > GetMacroValue(main,RPMcommand)
>  > cfpopen(/bin/rpm -q --queryformat "%{EPOCH}:%{VERSION}-%{RELEASE}\n"
>  > ntp)
>  > Appending [(none):4.2.0-7]
>  > cfpclose(pp)
>  > cfpopen - Waiting for process 23981
>  > RPM Package ntp not installed.
>  > AddMultipleClasses(ntp_uninstalled)
>  > 
>  > So looking at the code this seems to indicate that the rpm command is
>  > succeeding but the cfpclose. 
>  > 
>  > And now I've found
>  > http://lists.gnu.org/archive/html/bug-cfengine/2005-02/msg00013.html
>  > which describes the problem perfectly and there's a patch attached I'm
>  > going to try. But this sounds like the fix, and I can confirm the
>  > problem is still there in 2.1.13.
>  > 
>  > On Wed, 2005-03-09 at 13:35 -0500, Eric Dorland wrote:
>  > > Hello,
>  > > 
>  > > I'm attempting to roll out cfengine to a bunch of our servers (and more
>  > > to come). I'm finally grokking everything, but I've run into one last
>  > > problem I can't seem to solve. I've got a setup with all my master files
>  > > are in a CVS repository. Cfagent pulls the files out of cvs whenever it
>  > > updates. I also have a script to run cfrun everytime I commit something
>  > > to the CVS repo. 
>  > > 
>  > > My problem is that cfagent is behaving differently when invoked from
>  > > cfrun than it's periodic runs from cfexecd. The periodic runs through
>  > > cfexecd work great. When I invoke cfrun however, the "packages" target
>  > > doesn't work. I have cfengine check to see if the ntp package is
>  > > installed. If it isn't it installs it using up2date. I know the ntp
>  > > package is installed, but whenever I invoke cfagent through cfrun the
>  > > packages target says it's not installed. When run through cfexecd
>  > > everything works as expected.
>  > > 
>  > > I hope that was clear. I'm running the latest 2.1.13 on RHEL3. Has
>  > > anyone seen a similar problem? (Googling didn't come up with anything
>  > > similar). Let me know any debugging output I could provide that would be
>  > > helpful. Thanks.
>  > > 
>  > 
>  > -- 
>  > Eric Dorland
>  > eric.dorland@mcgill.ca
>  > WSG
>  > 514.398-5023 ext. 09562
>  > 
>  > 
>  > 
>  > _______________________________________________
>  > Help-cfengine mailing list
>  > Help-cfengine@gnu.org
>  > http://lists.gnu.org/mailman/listinfo/help-cfengine
> 

-- 
Eric Dorland
eric.dorland@mcgill.ca
WSG
514.398-5023 ext. 09562