Time jump in syslog-ng log files

Question

I am running my proprietary application (the name of which I would prefer not to disclose) in Linux 2.6 kernel, running on PowerPC (P2020) in an embedded environment. I am using syslog-ng to collect the logs from my application.

When the system is heavily loaded during the running of the application, it is seen that there is a jump in the timestamps in the log files. ie, No logs in the log files for considerable amount of time.

I have enabled Watchdog hardware to reset the processor if the system hangs due to running the application but no reset of the processor is seen and so it means that the application is running fine. Yet still the time jump occurs when the system is heavily loaded.

Is this a known behavior of syslog-ng? Is there any way to avoid this time jump? Please give me some input or documentation where this problem is already reported and the solution given.

Answer 1

I think this is more integral than just any package like Syslog unfortunately. I know the internal Linux clock as the kernel clock but it has other names.

I had a lot of problems with high load on ancient SPARC64 systems running Gentoo which caused the clock to skew frequently. Even ntpd wasn't suitable in the end (it's a long story) so I ended up cronning ntpdate ntp.blah.com every 1/2hr to keep the clock mostly sane and then also wrote the time to the hardware clock at each boot using hwclock.

More here.

Searching for kernel clock, system clock linux or similar is probably the way to go if that's not enough detail in the mini HOWTO.