In trying to test the use of a private mesh-connected NTP server, I find am unable to get my node to access the other users IP address from the node where I am running a tunnel client and tunnel server. The NTP server's address is reachable from the other nodes on my mesh, as well as a laptop plugged into the tunnel server node, but not from the OS of the router running the tunnel software. I tried 3.1.15 and 3.1.16b01.
Also: Suggestion for the NTP feature introduced 3.1.16 - when the user types in an address and hits the save button, try to ping that address and report whether or not it is reachable. It may be that it is not reachable in the environment where the data is being entered, so this is just advice, not a show-stopper. But if the address is no good and it is supposed to be good, that would be useful information.
Ken
I bring this up in hopes that a user can set at least 2 serves to poll and have something to click to Sync the clock, and then get a pop up window or DIALOG confirming the server is reachable and the node is synced.
Honestly though it sounds like a lot more work to me than its worth however considering that mesh nodes are NOT TIME CRITICAL devices, even if you had an accurate to the picosecond time server the nodes are not going to need time that accurate, are probably not even that accurate internally, and just network latency alone could throw that accuracy out the window.
The nodes are not Visa/Mastercard transaction servers, they are routers, microsecond accuracy for internal systems usage (graphs) really isn't critical on them for the locations they will be deployed to.
Ntpd is designed to query a pool of servers. If the GUI would alow multiple ntp servers... so I don't have to go and add them to the config my self. I presented a level of accuracy that not everyone may want to maintain. I would want the nodes to compare to at least 2 sources, what ever they might be. Every default install I've seen of ntpd on Linux has a pool of 4 to compare too. Those don't even cut it for jt65. One or 2 seconds +/- is not OK for me, when the default pool drifts.
I ran in to the problem back when I got my first rpi and wanted to script cron jobs, on a pi that wasn't connected to the internet. So I've been using GPS and now pps to sync.
Thanks for the ntp feature!
More than likely our nodes won't typically have internet access, so I plan on setting up a dedicated NTP server on my AREDN node. What is the proper configuration to make sure nodes throughout the mesh are able to get the latest time from my NTP server? Will it just be automatic?
Can someone tell me where the config file is so I can manually add a second server?
Tried searching a bit and didn't see any likely suspects.
Cheers.
Nope.
You can add a local NTP server (ex. RPi) and configure its ntp.conf with multiple NTP servers.
Chuck