[meta] Collect time offset samples

MarcoFalke commented at 10:14 am on April 8, 2022: member

To evaluate how useful/useless adjusted network time is, it could make sense to collect samples?

Be aware that time offsets are expected to be off, as some peers will send the network time as their own time, see commit 14ba286556faad794f288ef38493c540382897cb. Also, offsets of 1sec (or a few secs) are expected due to second-precision rounding errors and network delays. Also, larger offsets are expected for some peers that did a -reindex, see commit 0ac42c49f7bfc1e158a83c6d98bc465b06894e8a.

MarcoFalke added the label Brainstorming on Apr 8, 2022

MarcoFalke added the label P2P on Apr 8, 2022

jonatack commented at 2:24 pm on April 8, 2022: member

A patch that adds getpeerinfo timeoffset to -netinfo, if helpful to anyone: https://github.com/jonatack/bitcoin/commits/netinfo-time-offset

MarcoFalke commented at 12:37 pm on April 21, 2022: member

For the data to be meaningful, it would ideally also include the ua-string

MarcoFalke commented at 12:38 pm on April 21, 2022: member

Going to close this, as this is not an actionable issue. Though, comments are still welcome.

MarcoFalke closed this on Apr 21, 2022

vasild commented at 1:01 pm on April 21, 2022: member

 0$ bitcoin-cli getpeerinfo | jq .[].timeoffset | sort -n | uniq -c
 1   1 -43374
 2   2 -1879
 3   1 -9
 4   4 -4
 5   5 -3
 6   5 -2
 7  21 -1
 8  53 0
 9   1 1
10   1 2
11   1 57
12   1 108
13   1 166

 0$ jq -r 'sort_by(.timeoffset) | map({timeoffset: .timeoffset | tostring, subver: .subver}) | map(.timeoffset + " " + .subver) | .[]'
 1-43374 /Satoshi:0.16.3/
 2-1879 /nodes.mom.market:0.2/
 3-1879 /nodes.mom.market:0.2/
 4-9 /Satoshi:22.0.0/
 5-4 /Satoshi:0.21.0/
 6-4 /dsn.tm.kit.edu/bitcoin:0.9.99/
 7-4 /Satoshi:21.99.0/
 8-4 /Satoshi:0.21.1/
 9-3 /Satoshi:0.14.0/
10-3 /Satoshi:0.14.0/
11-3 /Satoshi:22.0.0/
12-3 /Satoshi:22.0.0/
13-3 /Satoshi:22.0.0/
14-2 /Satoshi:0.20.99/
15-2 /Satoshi:22.0.0/
16-2 /Satoshi:22.0.0/
17-2 /Satoshi:22.0.0/
18-2 /Satoshi:22.0.0/Knots:20211108/
19-1 /Satoshi:0.21.0/
20-1 /seeder:0.0001/
21-1 /Satoshi:22.0.0/
22-1 /bitnodes.io:0.2/
23-1 /Satoshi:22.0.0/
24-1 /Satoshi:0.17.0.1/
25-1 /Satoshi:22.0.0/
26-1 /Satoshi:22.0.0/
27-1 /Satoshi:22.0.0/
28-1 /Satoshi:22.0.0/
29-1 /Satoshi:22.0.0/
30-1 /Satoshi:22.0.0/
31-1 /Satoshi:22.0.0/
32-1 /Satoshi:22.0.0/
33-1 /Satoshi:22.0.0(Bitch)/
34-1 /Satoshi:23.99.0(@dunxen)/
35-1 /Satoshi:22.99.0/
36-1 /Satoshi:22.99.0/
37-1 /Satoshi:22.0.0/
38-1 /Satoshi:0.16.0/
39-1 /Satoshi:22.0.0/
400 /Satoshi:0.18.0/
410 /seeder:0.0001/
420 /Satoshi:22.0.0/
430 /bitnodes.io:0.2/
440 /bitnodes.io:0.2/
450 /Satoshi:0.21.1/
460 /bitnodes.io:0.2/
470 /Satoshi:0.21.0/
480 /Satoshi:0.21.0/
490 /Satoshi:0.21.0/
500 /Satoshi:22.0.0/
510 /dsn.tm.kit.edu/bitcoin:0.9.99/
520 /bitnodes.earn.com:0.1/
530 /bitnodes.earn.com:0.1/
540 /dsn.tm.kit.edu/bitcoin:0.9.99/
550 /bitnodes.io:0.2/
560 /Satoshi:22.0.0/
570 /Satoshi:22.0.0/
580 /dsn.tm.kit.edu/bitcoin:0.9.99/
590 /dsn.tm.kit.edu/bitcoin:0.9.99/
600 /Satoshi:22.0.0/
610 /Satoshi:22.0.0/
620 /Satoshi:22.0.0/
630 /Satoshi:22.0.0/
640 /Satoshi:22.0.0/
650 /bitnodes.io:0.2/
660 /Satoshi:0.16.1/
670 /Satoshi:22.0.0/
680 /Satoshi:22.0.0/
690 /Satoshi:22.0.0/
700 /Satoshi:23.0.0/
710 /Satoshi:22.0.0/
720 /Satoshi:22.0.0/
730 /Satoshi:22.0.0/
740 /Satoshi:22.0.0/
750 /Satoshi:0.19.0.1/
760 /Satoshi:22.0.0/
770 /btcwire:0.5.0/bitmarkd-payment-lightclient:0.1.0/
780
790 /Satoshi:22.0.0/
800 /Satoshi:22.99.0/
810 /Satoshi:22.99.0/
820 /Satoshi:22.99.0/
830 /Satoshi:22.99.0/
840 /Satoshi:22.99.0/
850 /Satoshi:22.99.0/
860 /Satoshi:22.99.0/
870 /Satoshi:22.99.0/
880 /Satoshi:22.0.0/
890 /Satoshi:23.99.0(@jon)/
900 /Satoshi:22.0.0/
910 /Satoshi:23.99.0(@jon)/
920 /bitnodes.io:0.2/
931 /Satoshi:22.0.0(Samourai Dojo 1.12.1)/
942 /Satoshi:0.18.1/
9557 /Satoshi:22.0.0/
96108 /Satoshi:22.0.0/
97166 /Satoshi:0.19.0.1/

jonatack commented at 8:05 am on April 23, 2022: member

MarcoFalke commented at 8:38 am on April 23, 2022: member

Looks like the data so far shows that all offsets are within 5 minutes, or they are so large (12 h for the 16.3 node) that time adjustment won’t work, or they are from clients that likely do not use our time adjustment logic or code (mom.market), or we don’t know whether adjusted time had an effect (all outbound nodes with user agent Satoshi:22.x or earlier)

jonatack commented at 8:50 am on April 23, 2022: member

Looks like the data so far shows that all offsets are within 5 minutes, or they are so large (12 h for the 16.3 node) that time adjustment won’t work, or they are from clients that likely do not use our time adjustment logic or code (mom.market), or we don’t know whether adjusted time had an effect (all outbound nodes with user agent Satoshi:22.x or earlier)

Yes, that’s a good summary of what I’ve been seeing so far.

DrahtBot locked this on Apr 23, 2023

[meta] Collect time offset samples #24805