1 |
Francesco Talamona wrote: |
2 |
> On Thursday 06 January 2011, Dale wrote: |
3 |
> |
4 |
>> Francesco Talamona wrote: |
5 |
>> |
6 |
>>> On Wednesday 05 January 2011, Dale wrote: |
7 |
>>> |
8 |
>>>> Now on my old system, it would adjust the drift file and the |
9 |
>>>> adjustments would get smaller and smaller. On the new rig, as |
10 |
>>>> you can see it stays about the same. I would like it to get to a |
11 |
>>>> point where it doesn't have to sync so often. I read on the |
12 |
>>>> website where they are needing more servers to help with the load |
13 |
>>>> and I don't want to be one of the ones putting a load on it. |
14 |
>>>> |
15 |
>>> Maybe you copied over /etc/adjtime from the previous machine. I |
16 |
>>> would try to regenerate it... |
17 |
>>> |
18 |
>>> Ciao |
19 |
>>> |
20 |
>>> Francesco |
21 |
>>> |
22 |
>> I'm sure I didn't copy that. I copied ntp.conf but that is all. I |
23 |
>> didn't even notice that one being there. I got to see what purpose |
24 |
>> that has. |
25 |
>> |
26 |
>> I may delete it tho and see what happens. It would generate a new |
27 |
>> one if I restart the service correct? |
28 |
>> |
29 |
>> Dale |
30 |
>> |
31 |
>> :-) :-) |
32 |
>> |
33 |
> It makes the hardware clock take care of the systematic drift. If a |
34 |
> wrong value is stored in it, it can interfere with ntp in the way you |
35 |
> described: every time ntp runs it always corrects for the same amount. |
36 |
> |
37 |
> From man 8 hwclock: |
38 |
> |
39 |
> "The Hardware Clock is usually not very accurate. However, much of |
40 |
> its inaccuracy is completely predictable - it gains or loses the same |
41 |
> amount of time every day. |
42 |
> This is called systematic drift. hwclock's "adjust" function |
43 |
> lets you make systematic corrections to correct the systematic drift." |
44 |
> |
45 |
> and: |
46 |
> "It is good to do a hwclock --adjust just before the hwclock --hctosys |
47 |
> at system startup time, and maybe periodically while the system is |
48 |
> running via cron." |
49 |
> |
50 |
> This is my "recipe": |
51 |
> let ntpdate sync your clock, then /sbin/hwclock --systohc |
52 |
> and you are done. From that moment on ntp takes care of the non |
53 |
> systematic error, while the drift is zeroed "by hwclock". |
54 |
> |
55 |
> Somewhere it is suggested to run /sbin/hwclock --adjust once a year. |
56 |
> |
57 |
> I experienced what you describe when I built my new machine, I had |
58 |
> copied /etc/adjtime (without knowing what it was) from the previous, it |
59 |
> took me a good deal of googling... |
60 |
> |
61 |
> BTW I don't have ntp.drift, I only use ntpdate and the clock is always |
62 |
> correct. |
63 |
> |
64 |
> HTH |
65 |
> Francesco |
66 |
> |
67 |
> |
68 |
|
69 |
Well, I tried openntp for a good while last night and it was worse than |
70 |
ntp. It was adjusting the clock by something like 20 seconds at a |
71 |
time. Even ntp wasn't that bad. I went back to plain ntp. This is |
72 |
what ntp has sent to messages overnight while I was napping: |
73 |
|
74 |
Jan 6 00:13:21 localhost ntpd[23712]: ntpd 4.2.6p3@1.2290-o Thu Jan 6 |
75 |
05:41:55 UTC 2011 (1) |
76 |
Jan 6 00:13:21 localhost ntpd[23713]: proto: precision = 0.102 usec |
77 |
Jan 6 00:13:21 localhost ntpd[23713]: Listen and drop on 0 v4wildcard |
78 |
0.0.0.0 UDP 123 |
79 |
Jan 6 00:13:21 localhost ntpd[23713]: Listen and drop on 1 v6wildcard |
80 |
:: UDP 123 |
81 |
Jan 6 00:13:21 localhost ntpd[23713]: Listen normally on 2 lo 127.0.0.1 |
82 |
UDP 123 |
83 |
Jan 6 00:13:21 localhost ntpd[23713]: Listen normally on 3 eth0 |
84 |
192.168.2.5 UDP 123 |
85 |
Jan 6 00:13:21 localhost ntpd[23713]: Listen normally on 4 eth0 |
86 |
fe80::1e6f:65ff:fe4c:91c7 UDP 123 |
87 |
Jan 6 00:13:21 localhost ntpd[23713]: Listen normally on 5 lo ::1 UDP 123 |
88 |
Jan 6 00:13:21 localhost ntpd[23713]: peers refreshed |
89 |
Jan 6 00:13:21 localhost ntpd[23713]: Listening on routing socket on fd |
90 |
#22 for interface updates |
91 |
Jan 6 09:09:31 localhost ntpd[23713]: Attempting to register mDNS |
92 |
Jan 6 09:09:31 localhost ntpd[23713]: *** WARNING *** The program |
93 |
'ntpd' uses the Apple Bonjour compatibility layer of Avahi. |
94 |
Jan 6 09:09:31 localhost ntpd[23713]: *** WARNING *** Please fix your |
95 |
application to use the native API of Avahi! |
96 |
Jan 6 09:09:31 localhost ntpd[23713]: *** WARNING *** For more |
97 |
information see <http://0pointer.de/avahi-compat?s=libdns_sd&e=ntpd> |
98 |
Jan 6 09:09:31 localhost ntpd[23713]: mDNS service registered. |
99 |
root@fireball / # |
100 |
|
101 |
It doesn't show that it is doing anything but the clock is somewhat |
102 |
accurate but it is still having to adjust it like before. I use ntpdate |
103 |
to see how far the clock is off. If I run it every few minutes, I can |
104 |
see it drift further away from the correct time then when ntp syncs, it |
105 |
resets it again. It did generate the ntp.drift file but it is still not |
106 |
adjusting the clock like on my other rig. |
107 |
|
108 |
I also read where the caps USE flag should be enabled. I did but it |
109 |
doesn't appear to have helped any. |
110 |
|
111 |
I deleted the /etc/adjtime but it has not been recreated. What creates |
112 |
this file? I deleted it a couple days ago and nothing has brought it |
113 |
back yet. |
114 |
|
115 |
I'm about ready to see what Linux would be like with no freaking clock |
116 |
at all. lol Now to see what I am going to try next. |
117 |
|
118 |
Dale |
119 |
|
120 |
:-) :-) |