ntpMerlin ntpMerlin v3.x

[pdc]

Hi all, I just set up ntpMerlin a few days ago, and am seeing an oddly cyclical pattern to Drift. Any idea what might be going on here?

I think the numbers are probably just fine (~300 nanoseconds Offset and ~300 parts per billion Drift!) but it is strange. For some reason it drifts low every night, with the maximum low at 4:30am. A random thought -- given how tiny the numbers involved are, might this be related to room temperature?!? (It's around 68 in the day and 62 at night) Though the CPU is generally around 70 °C whenever I've checked (i.e. during the day, not so much at 4am ). This is an AC86U running 384.19.

Thanks @Jack Yaz for all the cool software! (It's not an addiction, I can quit any time )

 

[heysoundude]

Hi all, I just set up ntpMerlin a few days ago, and am seeing an oddly cyclical pattern to Drift. Any idea what might be going on here?

I think the numbers are probably just fine (~300 nanoseconds Offset and ~300 parts per billion Drift!) but it is strange. For some reason it drifts low every night, with the maximum low at 4:30am. A random thought -- given how tiny the numbers involved are, might this be related to room temperature?!? (It's around 68 in the day and 62 at night) Though the CPU is generally around 70 °C whenever I've checked (i.e. during the day, not so much at 4am ). This is an AC86U running 384.19.

Thanks @Jack Yaz for all the cool software! (It's not an addiction, I can quit any time )

ha...sure you can (just like the rest of us ;-p )
zoom out - look at the 30d
However - also look at the sources you're referencing in the .conf : the closer to you and your ISP's server, the better, but the more often you ping a stratum 1 server, the smoother your charts will be.
from the link in an earlier post, you could add maxpoll to any of the servers you specify - maxpoll 7 minpoll 5 should get you pinging every 64 sec - a single routing device doing that shouldn't get the maintainers as irked a a swarm of devices coming from the same IP addy (but you should check their fair use statements to make sure)
Temp variations may have a small effect on things...but it's more likely traffic between you and your ntp sources that's causing the variations.
hows your QoS setup? is it as fine tuned as you think it can be, if youre running one?

HTH

 

[Vexira]

just wanted to ask if there is any way to fix this message "Missing NTS support" other than to turn off the cloudflare server.

 

[Wade Coxon]

Hi all, I just set up ntpMerlin a few days ago, and am seeing an oddly cyclical pattern to Drift. Any idea what might be going on here?

It could very well be temperature related since the drift absolutely follows a normal day's temperature curve.

It is also possible that there are other upstream latency related factors at play that cause your drift to be measured incorrectly.

Your baseline drift is already so low, that any minor environmental effect will be seen. It is very unusual for a consumer device's oscillators to measure so low in my experience. (my router hovers around 10ppm +/- 2ppm).

Out of interest, try pointing a fan at the router for a day, and see if that noticably affects the daytime drift.

It's definitely worth investigating out of pure interest, but absolutely nothing to worry about at all.
There is virtually no application you would need in a home setup that would benefit from getting your time any more precise .

 

[ottofreud]

It could very well be temperature related since the drift absolutely follows a normal day's temperature curve.

It is also possible that there are other upstream latency related factors at play that cause your drift to be measured incorrectly.

Your baseline drift is already so low, that any minor environmental effect will be seen. It is very unusual for a consumer device's oscillators to measure so low in my experience. (my router hovers around 10ppm +/- 2ppm).

Out of interest, try pointing a fan at the router for a day, and see if that noticably affects the daytime drift.

It's definitely worth investigating out of pure interest, but absolutely nothing to worry about at all.
There is virtually no application you would need in a home setup that would benefit from getting your time any more precise .

Just to chime in, I see that same "saw tooth" pattern in my graphs (in all three 30, 7, 1 day views). I have a fan on my AX88U and the CPU temp stays between 37C and 40C pretty constantly throughout the day (with router use, i.e. me staring at all the pretty graphs, being the factor that drives temps towards 40C) I use 0.us.pool.ntp.org, 1.us.pool.ntp.org, 3.us.pool.ntp.org as my servers. I have not played with Maxpool, Minpool settings. My avg drift sits around 4.4 ppm. I use Chrony since my ooma telo device will not sync to the ooma network if I use NTP (I don't even have a guess as to why). I was going to post a similar question about that interesting oscillation in the drift graphs. So now that the question is on the table, I thought I'd add some of my data points to the discussion.

 

[bluepoint]

just wanted to ask if there is any way to fix this message "Missing NTS support" other than to turn off the cloudflare server.

Just remove the NTS directive from cloudflare. You are getting the info because the server doesn't support NTS.

 

[Wade Coxon]

the CPU temp stays between 37C and 40C pretty constantly throughout the day

Still worth trying the fan as the drift may not follow the CPU temperature, since the clock signal could be coming from another component external to the CPU.

 

[ottofreud]

Still worth trying the fan as the drift may not follow the CPU temperature, since the clock signal could be coming from another component external to the CPU.

yes, I have a fan on mine as stated. Without the fan, the CPU temps are in the 55C range.

 

[Wade Coxon]

yes, I have a fan on mine as stated. Without the fan, the CPU temps are in the 55C range.

Sorry, I should have been more clear. Try the opposite case...without the fan, and see if there is a difference.

 

[ottofreud]

Sorry, I should have been more clear. Try the opposite case...without the fan, and see if there is a difference.

Good test. Won't be able to do that immediately, but will do so in the next day or so for 12 - 24 hours and see what the results are. I'll report back when I get some data.

 

[heysoundude]

It could very well be temperature related since the drift absolutely follows a normal day's temperature curve.

It is also possible that there are other upstream latency related factors at play that cause your drift to be measured incorrectly.

Your baseline drift is already so low, that any minor environmental effect will be seen. It is very unusual for a consumer device's oscillators to measure so low in my experience. (my router hovers around 10ppm +/- 2ppm).

Out of interest, try pointing a fan at the router for a day, and see if that noticably affects the daytime drift.

It's definitely worth investigating out of pure interest, but absolutely nothing to worry about at all.
There is virtually no application you would need in a home setup that would benefit from getting your time any more precise .

Unusual? Mine is at 3.2ppm or lower.
The spike was (IIRC) the update to 386.1_2 from .1 - it flatlined when I switched to a stratum1 ntp reference source hosted by the IX where my ISP houses its servers. my ping is <7ms and jitter <1ms
I'm beginning to think these scripts are helping us see the flaws in mass production, and how variances within component tolerances compound

 

[JohnD5000]

How much data is used each time ntpMerlin runs a check? Looking at the chart, it looks like it runs about 6x per hour or 144 times per day. I have it set to use a chrony timeserver if that matters. Is the frequency of how often ntpMerlin runs configurable someplace?

I'm asking because I use Comcast and they have announced that they are going to start charging us when we go over a certain amount of data per month.

 

[Wade Coxon]

How much data is used each time ntpMerlin runs a check? Looking at the chart, it looks like it runs about 6x per hour or 144 times per day. I have it set to use a chrony timeserver if that matters. Is the frequency of how often ntpMerlin runs configurable someplace?

I'm asking because I use Comcast and they have announced that they are going to start charging us when we go over a certain amount of data per month.

The data is miniscule. You are looking at something like tens of kB a day per server you are checking against.

It's best to let Chrony manage the poll rates, as these will adjust dynamically depending on how severe your drift and availability is.

 

[Maverickcdn]

yes, I have a fan on mine as stated. Without the fan, the CPU temps are in the 55C range.

(shameless plug)

You could install tempmon and reference the logged temp from syslog vs the logged drift

 

[Jack Yaz]

(shameless plug)

You could install tempmon and reference the logged temp from syslog vs the logged drift

What, no charts?

 

[JohnD5000]

The data is miniscule. You are looking at something like tens of kB a day per server you are checking against.

It's best to let Chrony manage the poll rates, as these will adjust dynamically depending on how severe your drift and availability is.

Thanks, can continue to run ntpMerlin then without worrying about data load when Comcast institutes their new policy

 

[Vexira]

Just remove the NTS directive from cloudflare. You are getting the info because the server doesn't support NTS.

im just wondering if support is even possible

 

[Jack Yaz]

im just wondering if support is even possible

not yet. Ask these guys https://github.com/Entware/Entware/issues/565

 

[chongnt]

What, no charts?

Is this a hint something new is coming soon?

 

[Jack Yaz]

Is this a hint something new is coming soon?

from me? no
but if @Maverickcdn wants to add historical logging and charts, I'm happy to lend a hand if required