ntpMerlin ntpMerlin - NTP Daemon for AsusWRT Merlin

[^Tripper^]

Noticed something interesting in my setup with an 87U. I’m running the usual scripts (diversion, skynet) and have some IOT thingys in my place. I’ve block internet access to and from the IOT gear and as such they don’t update time but I’ve also noticed that my normal internet surfing page loads would slow down tremendously after about 24-48 hours, almost as if the router was struggling with something if the IOT gear was blocked. A reboot usually sorts it out. Never could figure out why.

With this script running , I’ve had zero slowdowns so far and my IOT gear is now updating time perfectly.

Related? Who knows, I haven’t the foggiest idea but all I know is that since I’ve started using this, my network has not been slowing down.

Thank you for sharing this!

 

[Jack Yaz]

Isn't it a cron job that runs every 5 minutes?

Correct. The graphs are stored in a tmp folder that is accessible by the webui, so are lost on reboot,so won't be generated until the next 5 minutes rolls around. Perhaps I should add a call to generate them to s77ntpd

 

[Jack Yaz]

While my modem was messing around yesterday, I noticed issuing

service restart_wan

seemed to revive the built-in ntp client. While I could attach to service-event, it's possible there are other triggers that could awake ntp, so I wonder if it would be prudent to add

killall ntp

to the cron job (stat generation)?

@JDB @EmeraldDeer thoughts?

 

[EmeraldDeer]

While my modem was messing around yesterday, I noticed issuing

service restart_wan

seemed to revive the built-in ntp client. While I could attach to service-event, it's possible there are other triggers that could awake ntp, so I wonder if it would be prudent to add

killall ntp

to the cron job (stat generation)?

@JDB @EmeraldDeer thoughts?

Yes, I agree that adding "killall ntp" to "/jffs/scripts/ntpmerlin generate" is the way to go.

 

[Billy Chaney]

Correct. The graphs are stored in a tmp folder that is accessible by the webui, so are lost on reboot,so won't be generated until the next 5 minutes rolls around. Perhaps I should add a call to generate them to s77ntpd

Yes when I press the refresh button after a reboot, the graph breaks. I hope this helps in finding a solution.

 

[Billy Chaney]

@Jack Yaz Last upgrade seems to have fixed it. I'll keep an eye on it. Thank you very much.

 

[QuikSilver]

Any reason why info is shown in the offset graph (min: -1.1 m Max 701.3 u Curr: 701.3 u) Nothing shows at all for jitter though.....Whats "u" stand for btw?

 

[Jack Yaz]

Any reason why info is shown in the offset graph (min: -1.1 m Max 701.3 u Curr: 701.3 u) Nothing shows at all for jitter though.....Whats "u" stand for btw?

How many servers have you got configured in ntp.conf?

u stands for microseconds, I believe

 

[cmkelley]

How many servers have you got configured in ntp.conf?

u stands for microseconds, I believe

Yep. https://physics.nist.gov/cuu/Units/prefixes.html

 

[Jack Yaz]

Yep. https://physics.nist.gov/cuu/Units/prefixes.html

Been a while since that Physics A Level

Goes to show if you don't use knowledge, you forget it!

 

[JDB]

How many servers have you got configured in ntp.conf?

u stands for microseconds, I believe

Technically it’s a ‘Mu’ µ


Sent from my iPhone using Tapatalk

 

[QuikSilver]

How many servers have you got configured in ntp.conf?

u stands for microseconds, I believe

just one for now...

 

[Jack Yaz]

just one for now...

Single server users seem to be getting 0 jitter - I don't know enough about ntpd to know if that's expected though, sorry

 

[Zonkd]

Single server users seem to be getting 0 jitter - I don't know enough about ntpd to know if that's expected though, sorry

Sounds like this might be normal... if jitter is measurement of difference between system time and ntp server, then surely if they’re in sync jitter would be zero. A system could easily keep time with a single ntp server where latency/routing to it remains virtually constant. Jitter would have to be higher if using 2+ ntp servers because of variable latency and no server being perfectly in sync.

https://tools.ietf.org/html/draft-ietf-ntp-bcp-02

 

[JDB]

Sounds like this might be normal... if jitter is measurement of difference between system time and ntp server, then surely if they’re in sync jitter would be zero. A system could easily keep time with a single ntp server where latency/routing to it remains virtually constant. Jitter would have to be higher if using 2+ ntp servers because of variable latency and no server being perfectly in sync.

https://tools.ietf.org/html/draft-ietf-ntp-bcp-02

This was my initial thought too, but add into the mix the variable latency of the internet and you are bound to get some tiny amount of jitter, single figure microseconds at least!
I can believe it on a local LAN-LAN sync for sure tho.


Sent from my iPhone using Tapatalk

 

[heysoundude]

Sounds like this might be normal... if jitter is measurement of difference between system time and ntp server, then surely if they’re in sync jitter would be zero. A system could easily keep time with a single ntp server where latency/routing to it remains virtually constant. Jitter would have to be higher if using 2+ ntp servers because of variable latency and no server being perfectly in sync.

https://tools.ietf.org/html/draft-ietf-ntp-bcp-02

right, there always has to be ONE master clock for everything to reference/lock to; a pool takes time to achieve consensus, and if you're waiting on that, you'll drift away. (I'm going to have to dig up some documents/links on the SMPTE time code I'm <ahem> familiar with -NTP protocol is based on the same or similar maths no doubt- to try and to get everyone on the same page as to how this stuff works...visual representations were key for me, but now to go back and find them for everyone)

I haven't installed this script yet, but I have changed the ntp reference in the router's GUI to Canada's NRC clock in Ottawa. Conveniently, our national broadcaster (CBC) relays a reference of that over the air every day at 1pm. I get no small amusement from hearing the beep marking 1pm end and seeing the clock in my desktop's toolbar or my phone simultaneously roll over from 12:59 to 13:00...as it should.

what I'd like to know is where is the atomic clock that all the rest reference? do they have an ntp server I can use?? Britain/Greenwich maybe?

 

[L&LD]

right, there always has to be ONE master clock for everything to reference/lock to; a pool takes time to achieve consensus, and if you're waiting on that, you'll drift away.

I haven't installed this script yet, but I have changed the ntp reference in the router's GUI to Canada's NRC clock in Ottawa. Conveniently, our national broadcaster (CBC) relays a reference of that over the air every day at 1pm. I get no small amusement from hearing the beep marking 1pm end and seeing the clock in my desktop's toolbar or my phone simultaneously roll over from 12:59 to 13:00...as it should.

what I'd like to know is where is the atomic clock that all the rest reference? do they have an ntp server I can use?? Britain/Greenwich maybe?

I don't think we (as in average joe) will ever get access to the atomic clocks used all around the world as the 'reference').

But can you imagine the fun we could have, though?

 

[heysoundude]

I don't think we (as in average joe) will ever get access to the atomic clocks used all around the world as the 'reference').

But can you imagine the fun we could have, though?

they're generally funded by the taxpayers. I would assume in the US, the Navy would have one accessible; in every country, there should be ONE that some branch of the government maintains...

 

[L&LD]

they're generally funded by the taxpayers. I would assume in the US, the Navy would have one accessible; in every country, there should be ONE that some branch of the government maintains...

The taxpayers fund everything, regardless of government mumbo-jumbo to the contrary.

 

[EmeraldDeer]

If NTP over the Internet can only synchronize to the single milliseconds at best, then what would be the benefit of your access to an atomic clock?

For example, you would have better time with a GPS connected NTP server on your LAN than you would connecting to an atomic clock somehow connected to an NTP server on the Internet that is tens or hundreds of milliseconds away.

The next steps would be GPS pulse per second clock discipline or Precision Time Protocol hardware timestamping on network packets.