Re: detecting failed clients and perl.

[David Ressman]

Thus spake Mike Lang (mlang@lanl.gov):
> I could see a mail filter that would keep a list of clients and check them 
> off as they run,
> or have cfengine touch a file on an nfs mount to show the last run.  I'm 
> suure someone has a
> better idea.

We don't really haver a better idea, just a different one.  Since our 
cfengine modules are already using a database connection, we just have 
a module that updates a table that keeps a timestamp of when each record 
was last updated.  It's pretty much the exact same thing you do with the 
nfs mounts, but we weren't using nfs.  15 minutes after our cfengine run
ends, we just run a simple script to check that database for hosts that
haven't run in more than an hour.

> Something completely different; cfwrap and cfmail use perl, do you keep a
> copy of perl local
> to each machine?  If I keep perl on an nfs mount this could be a point of
> failure since the cron
> jobs would fail.


The only problem with that touches on your second question.  Since we rely
on a perl module (although it could be any scripting language w/ DB 
support) to update this table, it means that not only do we have to have
a script interpreter on each client system, we have to have a database
module in each system's interpreter.

Since we were going to use perl+DBI on each system anyway, it's not a
problem for us.  We could get around this by making a module with database
support in a compiled language and distributing the binary as our module,
but we ended up deciding not to.

On an unrelated note, cfwrap and cfmail are uncomplicated enough that you
could certainly rewrite them with relative ease in Bourne shell for
maximum portability.

David