On Fri, Apr 01, 2005 at 04:45:50PM -0500, Jason A. Smith wrote:
> A long time ago, someone noticed that ganglia's sys_clock metric was
> basically the same as the gmond_started:
> 
> http://sourceforge.net/mailarchive/message.php?msg_id=5320970
> 
> >From the 2.x sources it appears that this metric was evaluated only
> once, at startup and then rebroadcast every 20-30 minutes.  I still see
> this metric in the 3.x sources but it doesn't appear to be included when
> gmond is running, it appears to be left out of the default gmond config.
> Was this a simple oversight?  Has the old problem of it never changing
> been fixed?

This metric should die.  It's nothing to a rather odd waste of resources
since there's no real point in knowing what time a machine thought it
was as some unknown time in the past.

If anything related to time is going to be logged, it should be the
offset from their primary NTP server and possiably counter frequency as
reported by "ntpdc -c loopinfo" or equivalent.

-- Brooks

-- 
Any statement of the form "X is the one, true Y" is FALSE.
PGP fingerprint 655D 519C 26A7 82E7 2529  9BF0 5D8E 8BE9 F238 1AD4

Attachment: pgpcPeshP2k2E.pgp
Description: PGP signature

Reply via email to