0

Today I decided to update my old el5 kernel with yum after years. What I didn't know is that the kernel that was running was custom and after updating and rebooting the network drivers started to fail loading and so I decided that booting from an older kernel would spare me more time than trying to fix the custom one.

The previous, custom kernel was 2.6.39 and the current, downgraded one is 2.6.18-308.11.1 (both x64).

The network drivers are loading correctly again after the downgrade, but now ping is not measuring in the microseconds range anymore. Instead, it'll only round up/down to the nearest milisecond integer, like this:

PING 10.10.10.1 (10.10.10.1) 56(84) bytes of data.
64 bytes from 10.10.10.1: icmp_seq=1 ttl=64 time=1.00 ms
64 bytes from 10.10.10.1: icmp_seq=2 ttl=64 time=0.000 ms
64 bytes from 10.10.10.1: icmp_seq=3 ttl=64 time=2.00 ms
64 bytes from 10.10.10.1: icmp_seq=4 ttl=64 time=1.00 ms
64 bytes from 10.10.10.1: icmp_seq=5 ttl=64 time=0.000 ms

Is there any known settings in the kernel that makes this happen? I searched a bit on Google but didn't find a single individual with the same problem.

10
  • 2
    I'm not sure why that would even matter since ping itself isn't really that accurate. The margin of error for ping is really in milliseconds, and it increases with distance. Anything in the range you show matters not a bit to any real application.
    – Ron Maupin
    Oct 28, 2015 at 23:47
  • This server is used for network monitoring and now all the graphs are very hard to read. Previously: puu.sh/l1krf/78cd380e88.png now: puu.sh/l1kvB/e1d23c3973.png
    – Railander
    Oct 28, 2015 at 23:52
  • Microseconds in ping times are meaningless data, Gathering meaningless data is a waste of time. Graphing meaningless data just produces meaningless graphs. You can re-jigger the graphs to line up correctly with the current data. You should have probably just dropped the microseconds on the original. What really matters for ping times is probably closer to 10 milliseconds, and, if I were doing this, I would round to the nearest 10 milliseconds and graph that.
    – Ron Maupin
    Oct 29, 2015 at 0:04
  • 1
    Honestly ping isn't anywhere near accurate enough for it to matter.
    – Daniel
    Oct 29, 2015 at 0:22
  • @Daniel, no kidding, but I have been around long enough to know that there are thick-headed management-types who thrive on this sort of thing. It's no use arguing with them, but maybe, just maybe, you can talk them into something a little more reasonable. I have had managers who would have had coronaries if faced with this situation. Hence my suggestion of rounding to the nearest 10 milliseconds.
    – Ron Maupin
    Oct 29, 2015 at 1:31

1 Answer 1

0

Unfortunately, it looks like high resolution timers (sub millisecond) didn't fully land in the kernel until 2.6.21.

You'll probably have to re-upgrade your kernel to something released within the last 8.9 years. :-)

See http://elinux.org/High_Resolution_Timers or Google for "Linux high resolution timers", or check the time(7) man page.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .