1 |
On 03/17/2016 02:03 PM, Bill Kenworthy wrote: |
2 |
> On 17/03/16 20:26, Alan McKinnon wrote: |
3 |
>> On 17/03/2016 08:50, Håkon Alstadheim wrote: |
4 |
>>> I have a server SUPPOSED to be running 24/7, but every once in a while |
5 |
>>> during a prolonged absence the box will go down. The Real Time Clock |
6 |
>>> will drift, and in the rush to get the box up again I let everything |
7 |
>>> boot up automatically and get both wrong time on the main systems, and |
8 |
>>> different times on the various systems. |
9 |
>>> |
10 |
>>> My setup has a main server which does NTP, but with no direct link to |
11 |
>>> the outside. Router&firewall /have/ to be booted booted later (dumb |
12 |
>>> setup, don't ask), after which I can finally get correct time from NTP. |
13 |
>>> |
14 |
>>> NTP initiates "11 minute mode", which makes /etc/adjtime useless as far |
15 |
>>> as I understand. Anybody have a /correct/ way to account for RTC drift |
16 |
>>> on a box running ntpd? Right now I have a ---file in |
17 |
>>> /etc/cron.d/time-bad like so: |
18 |
>>> * * * * * root adjtimex -S 5 >/dev/null 2>&1 </dev/null |
19 |
>>> --- |
20 |
>>> |
21 |
>>> Combined with an old-fashioned setup for hwclock during boot and |
22 |
>>> shutdown. This feels really wrong, and I have no idea what I am doing. |
23 |
>>> |
24 |
>>> TLDR: Anybody have a /correct/ way to account for RTC drift on a box |
25 |
>>> running ntpd? |
26 |
>>> |
27 |
>>> |
28 |
>> |
29 |
>> When the box was off, all questions of accurate ntp tracking are moot. |
30 |
>> ntp is designed around the idea that every second happens but from your |
31 |
>> machine's point of view they didn't happen since it was powered down. |
32 |
>> |
33 |
>> I would go the really simple route and force ntpdate to run once during |
34 |
>> boot up before ntpd is started, thereby avoiding the entire issue. |
35 |
Why can't I have proper drift information for my RTC ("bios clock") ? |
36 |
The old way ( where "system-time as set by ntp" minus "RTC time" |
37 |
gives "drift-value" written to /etc/adjtime ) used to work perfectly |
38 |
for me for several years. Is there no canonical way of getting that |
39 |
these days? |
40 |
|
41 |
My problem is that my WAN connection can not be brought up until well |
42 |
after the main server is up (stupid I know, but rearranging things |
43 |
entails a major overhaul). Thus a bios clock without drift information |
44 |
gives me a choice between ntpdate (which messes up my logs) and ntp with |
45 |
incremental adjustments (which might leave clocks wrong for several days). |
46 |
|
47 |
I really need the logs to be on the same clock for all systems. Don't |
48 |
ask, just assume I know why it's called bleeding edge . I also really |
49 |
need sub-minute accuracy on all clocks. I suppose I should try running |
50 |
ntpdate on everything once the WAN connection is up, just to see how bad |
51 |
the mess is. |
52 |
|
53 |
|
54 |
>> Sometimes correctness really doesn't matter, this looks like one of those. |
55 |
>> |
56 |
>> |
57 |
>> alan |
58 |
>> |
59 |
> add a cheap gps setup as the reference clock to the server, |
60 |
That sounds like a real option. I have an old Nokia N900 lying around |
61 |
with a broken usb-port, so I'd need to solder in a power lead. Any |
62 |
pointers for how to read time signal from the gps on a maemo system? |
63 |
> or even |
64 |
> better is a dedicated time server (either a real one or a raspberry |
65 |
> pi/gps) on the network if you have internal connectivity. Going super |
66 |
> cheap, but not quite as accurate for me was an arduino and rtc on a |
67 |
> bluetooth pan for when the network was down but I needed a reference (to |
68 |
> power up the real server :). |
69 |
> |
70 |
> |
71 |
> google "arduino time server" for plenty of options :) |
72 |
> |
73 |
This led me to finding some serial-port connected gps modules. |
74 |
Serial-ports I have, so this is sounding good. 35 to 40 euro for a gps |
75 |
device when hwclock and /etc/adjtime should give ballpark-correct time |
76 |
on boot makes me hate this though. |
77 |
|
78 |
This reminds me one reason I need a valid time is my DVB-T TV-receiver |
79 |
card. It should be possible to find a clock source in the broadcast |
80 |
stream. I'll research that first, while I leave my "adjtime -S 5 " hack |
81 |
running, even though I still don't know if that makes any sense at all. |
82 |
At least now there is something different from "0.0000" in the |
83 |
driftvalue, which gives me some hope I'm on to something. |