Mailinglist Archive: opensuse (4570 mails)

< Previous Next >
Re: [SLE] NTP set up in YAST is loosing time
  • From: "Carlos E. R." <robin1.listas@xxxxxxxxxx>
  • Date: Fri, 18 Nov 2005 13:11:50 +0000 (UTC)
  • Message-id: <Pine.LNX.4.61.0511181347540.8534@xxxxxxxxxxxxxxxx>
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1


The Thursday 2005-11-17 at 14:40 -0500, François Pinard wrote:

> > Many processes depend on timing. Queuing would go berserk for a while,
> > for instance.
>
> Of course, processes may depend on timing. But this is a generality...
>
> What were the symptoms of "queues going berserk" which you observed? Or if
> you did not observe them yourself, maybe you are sharing with us some
> experience you've got for someone else? Or you might have read the source
> code, and telling us your analysis?

It wasn't me who said "critical". Yes, I have observed weird things when
changing the time suddenly, specially if you change the date. If mail is
pending, for instance, and the time jumps forward ahead of the allowed
delay for delivery, the pending mail suddenly bounces back as
undeliverable. Yo might not need days for this: if postfix is sending
mails to a content filter, and it thinks they are taking more than half an
hour to process them, it aborts; it might then bounce or send without
filtering.

The mta should/might cope with the inconveniences of small jumps, but I
think you can understand that their queuing algorithms are thrown out of
balance. As for time going back... well, for instance, spammassassin will
note it and complain (DATE_IN_FUTURE_*). An email might be tagged as spam
because of that. Or, depending on the algorithm used, emails might stay
put till half an hour since the last try had passed, but as the time stamp
would be in the future, it could be hours till real delivery.

Or, if you are trying to analyze an event, something that went wrong, you
have a quite harder time tracking what happened and in what order if one of
the servers marks the time wrong. It could be "critical" depending on what
you are analyzing.

One thing you can easily observe is what happens with cron jobs. I tried.

- --
Cheers,
Carlos Robinson

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.0 (GNU/Linux)
Comment: Made with pgp4pine 1.76

iD8DBQFDfdLPtTMYHG2NR9URAolbAJ9LVbsr8Gc1p7skBEhU76xMCba2uACfdFLu
o3+dh2/0efj/jwUjYORFaiA=
=EuAb
-----END PGP SIGNATURE-----
< Previous Next >
Follow Ups