Nicolas Williams wrote:
>  - Where is sunman-stability used?

It isn't. Other SFW projects use it to add attributes on the end of 
the man pages, but I am supplying new man pages with the attributes 
section already included. I will remove it.

> 
>  - Is there a way to force quick timesync (a la ntpdate)?  Or must one
>    really way up to 1800 seconds in order to be assured that time is
>    synchronized (when config/wait_for_sync == true)?  

NTP v4 has a new feature such that adding the keyword "iburst" to a 
server line will cause that server to become usable in about 10 
seconds. Without iburst, it takes between 5 and 6 minutes.

> 
>    Looking at the NTP docs I see that the options that are relevant to
>    the above question are: -g (don't panic if offset > 1000s) and -q
>    (emulate ntpdate).
> 
>    That answers my earlier question about ntpdate.  We could have a
>    transient service to set the time quickly (ntpd -gq ...) and one to
>    run ntpd as a daemon.  ALTERNATIVELY you could have a property to
>    indicate that the start method should first run ntpd -gq then ntpd as
>    a daemon.

This is not needed. At anytime that ntpdate (or ntpd -gq) can get the 
the time from a remote server, so can ntpd without the -q. So, the 
best bet is to just start ntpd with the -g and iburst and after the 
first sync, let it run.

> 
>  - Be careful how you detect restarts.  Your current start method does
>    not detect this "svcadm stop -s ntp; svcadm start ntp".  You may want
>    to use a file in /var/run to track whether the clock has been set
>    once.

I have spent a good deal of time thinking about this. I wasn't sure if 
it was worth it because I kept coming up with corner cases. The 
current implementation always does the ntpdate, so this implementation 
  is more flexible than the current system.

How about having a property to determine whether or not to allow a 
single large offset other than at boot?  But how does the admin make a 
one time exception? I'd hate to have to have the admin change the 
property, start the service and then change it back. Deleting the file 
in /var/run would work too, but that is getting a little too deep into 
implementation details than I'd like.

> 
>  - The comment in the start method about "panic gate" could be clearer.
> 

Okay, I'll expand it.

>  - W.r.t. my earlier question re: privileges, ntpd needs proc_fork
>    (duh).
> 
> Nico

-- 
blu

"Mark my words, nanotechnology is going to be huge!"
----------------------------------------------------------------------
Brian Utterback - Solaris RPE, Sun Microsystems, Inc.
Ph:877-259-7345, Em:brian.utterback-at-ess-you-enn-dot-kom

Reply via email to