Re: cfengine horror stories

[Mark . Burgess]

Nate, 

I have never heard of this before. If I do this on my 2.0.4 version

tidy:

 //tmp r=inf age=0

cfagent -f ./cftest -d3|more

DEFINED TIDY MASKS

//tmp (maxrecurse = inf)
    pat=//tmp, a-age=0, size=0, linkdirs=k, rmdirs=n, travlinks=F compress=n
       Define
       ElseDefine
       recurse=inf

The parser interprets this correctly. However, when run, cfengine
interprets this as a wildcard for some reason...and the bug seems to
start there...ouch!! So, no, it is not fixed...


I shall look into this nasty bug and come up with a fix very soon,
Thank you for reporting this!

Mark



       

On 27 Nov, Nate Campi wrote:
> I was testing cfengine and grabbed a copy from sunfreeware.com, happened
> to be an old 1x version.
> 
> I introduced an error into a tidy section, so that /tmp became //tmp in
> the config file.  This caused cfengine to tidy the / filesystem,
> completely destroying the host.
> 
> I've since pushed out the latest 2x version via a sun package I built
> myself, but I only have the agent/cfexecd running on a few
> non-production hosts. I'm thinking to starting cfexecd live on all my
> hosts and slowly building up the configs over time, and only introducing
> changes to the production systems after I've tested on a staging
> cluster.
> 
> I think this will work fine, but I'm curious about two things:
> 
> 1) is the parser issue fixed so that tidying //tmp won't tidy / ?
> 2) are there any other horror stories anyone can share that might help
> me avoid any other stupid mistakes?



~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Work: +47 22453272            Email:  Mark.Burgess@iu.hio.no
Fax : +47 22453205            WWW  :  http://www.iu.hio.no/~mark
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~