sntpclock warnings and seldom failures on Gentoo Linux

[ydamyanov]

Anyone to have slightest idea what might be the cause of this bug:

http://bugs.gentoo.org/show_bug.cgi?id=112710

clockspeed has no Gentoo maintainer recently, that's why I am
bothering the clockspeed list.

Thanks and apologies,
Yassen

--

Yassen Damyanov
Troyer Information Systems

[Toby Betts]

Yassen Damyanov wrote:
> Anyone to have slightest idea what might be the cause of this bug:
>
> http://bugs.gentoo.org/show_bug.cgi?id=112710
>
> clockspeed has no Gentoo maintainer recently, that's why I am
> bothering the clockspeed list.

The messages you're seeing:

  sntpclock: warning: unable to read clock: bad response format
  sntpclock: fatal: time uncertainty too large

have more to do with your network and your choice of NTP server than
with the sntpclock software itself. The Network Time Protocol uses UDP,
a connectionless data transfer method with no built-in system of
congestion control. What this means is that sometimes UDP packets get to
their intended destination and sometimes they don't. Under very heavy
traffic, sntpclock won't he able to communicate with the NTP server in a
timely manner and thus won't have enough information to adjust the
system clock correctly. It will warn you and if things don't improve,
eventually die. Every time I've seen these messages, it's been because
of high traffic or a dodgy NTP server.


Toby

[Brian Lewis]

On Wed, 16 Nov 2005, Toby Betts wrote:
> The messages you're seeing:
>   sntpclock: warning: unable to read clock: bad response format
>   sntpclock: fatal: time uncertainty too large
> have more to do with your network and your choice of NTP server than
> with the sntpclock software itself.

I think this is probably right. I run Gentoo, and it seems to be working
alright.

$ sntpclock 130.184.7.93 | clockview
before: 2005-11-16 11:57:05.708563000000000000
after:  2005-11-16 11:57:05.703646797912597655

[ydamyanov]

brian wrote:

> On Wed, 16 Nov 2005, Toby Betts wrote:
>
>>The messages you're seeing:
>>  sntpclock: warning: unable to read clock: bad response format
>>  sntpclock: fatal: time uncertainty too large
>>have more to do with your network and your choice of NTP server than
>>with the sntpclock software itself.
>
>
> I think this is probably right. I run Gentoo, and it seems to be working
> alright.
>
> $ sntpclock 130.184.7.93 | clockview
> before: 2005-11-16 11:57:05.708563000000000000
> after:  2005-11-16 11:57:05.703646797912597655

Toby, Brian: thank you!

You are absolutely correct, so I have one stupid bug report to close.. ;-)

BTW, do you know a way to suppress those warnings?

Yassen

[Toby Betts]

Yassen Damyanov wrote:

> You are absolutely correct, so I have one stupid bug report to close.. ;-)
>
> BTW, do you know a way to suppress those warnings?

I wouldn't recommend it: if the software is having problems connecting,
I would want to know about it. Nonetheless, these warnings are simply
written to stderr, so redirecting stderr to /dev/null will hide them:

  $./sntpclock 1.2.3.4 2>/dev/null

Note that this will not fix anyone's problems, and now instead of
complaining about "annoying warning" bugs, they'll complain about "no
warnings when things go wrong" bugs. I'd rather explain the former than
do triage on the latter.


Toby