I have two boxes, on one box I get my ping times rounded to the nearest 4 ms, so I often get ping times of 4.00 ms or 0.000 ms. On the other box, which is almost identical, I get normal (correct) ping times, e.g. 3.16 ms, 1.41 ms From a hardware perspective, the box with the bad ping times has an extra processor and a bit more ram. From a configuration perspective, both boxes have the same .config file and no differing sysctl settings. CONFIG_HZ_250=y and CONFIG_HZ=250 are set (1000/4=250 so might be useful) The problem occurs on 2.6.23-gentoo-r9, 2.6.23-gentoo-r8. Downgrading to the previously used kernel, linux-2.6.19-gentoo-r5, resolves the problem. Reproducible: Always
What happens with 2.6.24-r3? If that also fails, can you try the latest development kernel which is vanilla 2.6.25-rc4 right now?
This is on a production box, and I can't reproduce it on other hardware. I can't install a development kernel on it, but I can certainly install the next production kernel when it comes out.
please post .configs from both systems i suspect one system has NO_HZ or hrtimer enabled and the other doesn't
Both servers have identical output from /proc/config.gz
please post the config anyway
Problem gone in 2.6.24-r3
Great, thanks for letting us know.