[ntp:questions] Ntpd in uninterruptible sleep?

A C agcarver+ntp at acarver.net
Mon Oct 24 20:54:49 UTC 2011


On 10/24/2011 11:48, unruh wrote:
> On 2011-10-24, A C<agcarver+ntp at acarver.net>  wrote:
>> On 10/24/2011 06:24, unruh wrote:
>>> On 2011-10-23, Uwe Klein<uwe at klein-habertwedt.de>   wrote:
>>>> A C wrote:
>>>>> More interesting is that the cpu was pegged until I was able to kill and
>>>>> restart ntpd.  Most of the cpu was devoted to ntpd during this locked up
>>>>> period.  Simple things like typing at the console were difficult.  It
>>>>> would take a few seconds for a keypress to register on the screen.  Once
>>>>> ntpd was restarted the system responded normally and the cpu usage
>>>>> dropped to normal levels.
>>>>>
>>>>> This is still version 4.2.6p3.  I should probably go ahead and compile
>>>>> the most recently released version but I'm at a loss to understand why
>>>>> it happened.
>>>>
>>>> CPU (over)loaded
>>>> or the system is swapping like mad ?
>>>
>>> That would destroy everything ( ie slow everything to a crawl). Also
>>> ntpd is a small program and especially at heightened niceness (which he
>>> said he used) should not get swapped out or affected.
>>>
>>>> ( I'd think it is swapping? )
>>>>
>>
>> Right, the swapping would only occur if I was trying to actively do
>> something while diagnosing the problem.  Otherwise the system load is so
>> low there's no real need to swap.  There is no desktop environment
>
> ? swapping occurs if system memory fills up completely and there is no
> more memory. It does not just refer to any disk access. Run top and see
> if there is a substantial swap useage. If there issomething is very
> wrong.

I was making the statement that swapping occured only when I was 
actively attempting to debug the situation.  Leaving the machine alone 
did not result in sustained swapping.

>
>> running on the system, only the standard daemons plus some extras (sshd,
>> ntpd, gpsd, crond, syslogd, inetd, postfix for system messages only) and
>> then one xterm (with ssh session inside) and one xclock.  The cron jobs
>> are mostly system housekeeping (log rotation, etc.) that occur at 0:00
>> but the crashes would occur at other times so none of the cron jobs were
>> running at the time of any crash.
>
> NOne of those would cause swapping unless you only have 1MB of memory.

I understand that, I was only documenting them for completeness.



More information about the questions mailing list