| 1 |
On Fri, Jun 08, 2007 at 10:38:00PM -0400, Richard Freeman wrote: |
| 2 |
> I just noticed that my time was off and after checking the logs I saw that ntpd was adjusting the time by 5 minutes several times a day for the last month. |
| 3 |
> |
| 4 |
> Searching around I found some hints that disabling apic might help. This is on a K8V deluxe motherboard and running 2.6.20-r7. Before I disable apic, will this have any negative |
| 5 |
> effects on the system? This is a desktop system so I don't really care about power-saving features (although I'd like to keep cpudyn working if possible). |
| 6 |
> |
| 7 |
> I'm also running ivtv (mythtv backend) which apparently can cause clock skew due to some kind of DMA error, but I'm not seeing that in the logs. |
| 8 |
> |
| 9 |
> Any advice? |
| 10 |
|
| 11 |
First, in general, both openntpd and ntpd are cranky little beasties. |
| 12 |
For a service that should be as basic and reliable as, say, ssh, an NTP |
| 13 |
installation requires a tremendous amount of ongoing TLC -- daemons |
| 14 |
randomly die or get wedged, or just fail to update the time correctly. |
| 15 |
Sometimes they actively skew a machine's time off. |
| 16 |
|
| 17 |
Second, the worst problems I ever saw with NTP weren't actually NTP's |
| 18 |
fault. I had a box which lost 15m/day, and (surprise surprise) NTP |
| 19 |
couldn't keep up and kept erroring out. It was an amd64 box that was |
| 20 |
(for reasons unbeknownst to me) installed x86, which I thought was the |
| 21 |
problem. Turns out it was bad RAM. The box ran for months as a |
| 22 |
postgresql server using that RAM, but NTP? No-can-do. Go figure. |
| 23 |
|
| 24 |
Dustin |
| 25 |
-- |
| 26 |
gentoo-amd64@g.o mailing list |
Archives