On 11 July 2013 20:59, Ashley Penney <ashley.pen...@puppetlabs.com> wrote:

>
> In the new code we set panic based on $is_virtual by default, so it sets
> panic to false for virtual and true for physical.
>

Indeed you do; I guess next time I take part in a discussion I should look
at *all* the code involved, and not just skim over stuff :-)



> That way we get the right behavior out of the box and physical people can
> override it too.  I figured that was preferable to having more logic in the
> templates.
>

Yep; plus the template wouldn't be able to generate a warning that a
user-defined option is going to be overridden, if that's the behaviour you
decide to implement.



> I suppose it depends on if there is the potential of a use case where
> people on virtual machines are simply not allowed to tolerate large skews
> either, I'd hate to railroad them by forcing the issue.
>

The only case I've come across where large steps are not
tolerated/permitted is in an Oracle RAC setup.  There, the '-x' option to
ntpd is required, and that forces it to always skew rather than step.

Assuming it's configured as per Oracle's and VMWare's recommendations, if a
VM-based RAC node is suspended then resumed some time later, it'll have a
large time difference to its clock source.  Given 'tinker panic 0', ntpd
will still be running, but given '-x' it will only slowly adjust the clock
forward.  At this point, RAC will evict the node due to the time
difference, causing the server to reboot.  On reboot, 'ntpdate' or 'ntpd
-q' will be run to set the clock to the correct time, and everything's back
to how it was before the VM was resumed.

Why did I mention all that? Well, in my opinion, that's about as harsh a
way to fix a large time difference as there is, and having ntpd panic or
not wouldn't have changed anything at all (RAC would still have rebooted
the box upon detection of the time diff).  Given all that, *I* can't think
of a scenario where you'd want ntpd to panic.  I'm always interested to
hear other's thoughts/opinions/scenarios though.

IMO, the ntp module should issue a notice() if both @panic and @is_virtual
are true; it's a bit more polite than just overriding someone's decision,
but might help them realise they're probably not doing the right thing.

Regards,

Matt.

-- 
You received this message because you are subscribed to the Google Groups 
"Puppet Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to puppet-users+unsubscr...@googlegroups.com.
To post to this group, send email to puppet-users@googlegroups.com.
Visit this group at http://groups.google.com/group/puppet-users.
For more options, visit https://groups.google.com/groups/opt_out.


Reply via email to