[18:03] and still https://paste.debian.net/plainh/f123e87f [18:06] however, https://status.canonical.com/ says "archive.ubuntu.com" is happy. [18:27] tomreyn: is that really a factor of 50 difference from the fastest to the slowest? [18:27] sarnold: yes [18:27] cripes [18:27] on this single not too reliable test [18:28] the script should be on github.com/tomreyn/scripts or something [18:29] i just ran it again, now it's just factor 25 [18:30] oh niiiiiice. I use one of yours all the time, I called it check_mirrors_last_modified_at_url .. somehow I don't remember this repo though :) [18:30] this stuff is completely buggy, dont expect too much ;) [18:30] I sure didn't expect curl -w :) I don't thikn I've seen that before [18:32] tomreyn: https://termbin.com/ucd5 [18:34] your routing is too good [18:34] cheater [18:35] you're probably itting next to that cdn, dont you? [18:35] * sitting [18:36] or knitting [18:36] hmm wait, you have varying response sizes [18:37] got a proxy? [18:55] hmm, now they're all fast here, too [18:56] but i have "response size (body): 25021109B" everywhere, not sure why yours would differ, sarnold [19:05] tomreyn: hey yeah why *are* they so much smaller??? [19:06] requesting that url with just curl http://archive.ubuntu.com/ubuntu/ls-lR.gz | wc shows more expected results, 99637 545603 25021109 [19:07] sarnold: this will print the commands run, and likely point to my mirtake: bash -x check_mirrors_speed_at_url http://archive.ubuntu.com/ubuntu/ls-lR.gz [19:08] tomreyn: I can't spot anything ;( https://termbin.com/t6a1 [19:10] hmm maybe it's a matter of http 1.1 vs 2 or something [19:10] i just finished a large project at work, not in the right mood to debug my scripts ;) but happily on the weekend [19:11] oo, congrats :) [19:12] hehe, thanks [19:21] tomreyn: i think it might be related to curl trying to write partial binary downloads to pipes. i can reproduce at https://pastebin.ubuntu.com/p/3vhY24dQhT/ but adding a "-o /dev/null" to curl seems to bring consistency to the requests [19:22] ah, thanks cmisare! i did push an update doing exactly that some minutes ago. [19:24] i guess i want --no-buffer [19:27] probably not, the idea is to the throw away the results :) [19:28] cmisare: awesome debugging :) [19:28] tomreyn: this is much more like it! https://termbin.com/g0lj [19:31] sarnold: much more consistent, nice. just the speed averages are printed in a different format for you, with 3 decimals [19:32] is this a different curl version? https://termbin.com/tb2a [19:32] or maybe it's your locale / shell [19:32] tomreyn: heh yeah, I thought ihad a good connection, but .3s to download ls-lR.gz sounded too good to be true .. but not enough for me to read it closely :) [19:33] curl 7.68.0 here (on focal) [19:33] LANG=en_US.UTF-8 [19:33] LC_TIME=C [19:34] LC_NUMERIC="en_US.UTF-8" + LC_MEASUREMENT="en_US.UTF-8" also? [19:35] yeah, probably they just inherit LANG [19:35] i have LC_TIME="en_US.UTF-8" (not C) but i think this applied to actuall wall clock time only. [19:35] I set mine to C because date -u started doing stupid things [19:36] like AM/PM? ;) [19:36] anyways, i think the original issue is no longer / not currently present, mirrors are doing fine. [19:37] thanks for debugging my script, though :) [19:37] woot!