2021-04-27T08:53:57 download.opensuse.org in USA is painfully slow 2021-04-27T08:54:24 has been for more than a day 2021-04-27T09:02:11 agony waiting on OSS repository retrieval 2021-04-27T09:48:55 a-865k: try http://mirrorcache-us.opensuse.org/download/ 2021-04-27T09:50:26 Eighth_Doctor: I'm pushing a lot of Factory package git repos atm. I found a much faster way to split them within seconds (compared to 5 minutes with git subtree split). 2021-04-27T09:52:21 and before that I found a way to break the pagure server by pushing repos that failed, but left data in /tmp until the partition was nearly full. 2021-04-27T10:56:32 bmwiedemann: I don't know whether to congratulate you or sigh in sadness 2021-04-27T10:57:44 probably worth fixing though ;) 2021-04-27T11:12:11 bmwiedemann: I've tried mirrorcache before, usually with very disappointing results. USA & Canada don't have enough mirrors of TW and repositories/ :( 2021-04-27T11:12:59 mirrorcache doesn't help for me either 2021-04-27T11:13:02 I just kinda give up 2021-04-27T11:13:46 a-865k: tw should not be an issue in the US, more than 2/3s of US mirrors carry TW. 2021-04-27T11:15:25 repositories - yeah, for most mirrors, it is just too big. 2021-04-27T11:16:54 pjessen: I stopped carrying TW because it never got used 2021-04-27T11:19:38 TW does seem less popular in some places. No mirrors in Canada, none in Australia. Still, 16 in the US isn't bad. 2021-04-27T11:20:55 yeah, the issue seems to be nobody actually gets redirected to those mirrors though, I hear people from even CA, not just NY are getting german mirrors instead of US 2021-04-27T11:22:48 LCP: it would be good to see some mirrorlist examples - for TW, there is really no reason for getting mirrors outside the US. 2021-04-27T11:23:12 indeed 2021-04-27T11:28:19 apart from that, download.o.o is pretty busy right now, 2021-04-27T11:28:45 a-865k, Conan Kudo: https://download.opensuse.org/tumbleweed/repo/oss/x86_64/zypper-1.14.43-1.2.x86_64.rpm.mirrorlist what do you see there 2021-04-27T11:29:14 I see 13 US mirrors 2021-04-27T11:29:53 several from kernel.org, two from provo-mirror, and 4 from universities far from me 2021-04-27T11:30:23 and bmw's mirror 2021-04-27T11:30:41 if you removed .mirrorlist from the end where does that redirect you? 2021-04-27T11:31:07 kernel.org this time 2021-04-27T11:31:47 hmmm 2021-04-27T11:35:58 yes, the provo mirror is counted twice, and I also counted Bernhards mirror :-) 2021-04-27T12:11:19 Sasi Olin: it's basically a roundrobin 2021-04-27T12:11:31 note that immediately out of the 13, the next mirror is from Japan 2021-04-27T12:11:53 hmmm 2021-04-27T12:13:14 basically what happens for me is that with a sufficiently big transaction, I wind up hitting one of these "bad" mirrors and everything gets stuck 2021-04-27T12:13:34 this happens on copr so frequently that we have retries encoded in the mock config 2021-04-27T12:13:56 I personally deal with it by using the metalink instead, because then DNF switches stuff around 2021-04-27T12:14:18 but the problem is that our mirroring doesn't work correctly for guaranteeing some degree of consistency for metalink fetching 2021-04-27T12:17:21 Eighth_Doctor: Japan?? that doesn't sound right, different continent. 2021-04-27T12:17:29 Yup 2021-04-27T12:18:10 which mirror is it? 2021-04-27T12:18:20 nrt.edge.kernel.org 2021-04-27T12:19:17 then a few NL ones, Swiss and German ones after, then tons of eastern Europe ones, and more Asian ones 2021-04-27T12:20:04 with nrt.edge, it has geo-location sensitive dns, when I just looked it up now, I got an address in new york 2021-04-27T12:21:01 interesting 2021-04-27T12:22:05 when there are 12 mirrors locally, I would not expect any redirects to mirrors in other countries. 2021-04-27T12:22:18 🤷‍♂️ 2021-04-27T12:24:48 s/countries/continents/ 2021-04-27T12:26:17 hm, I wonder if maybe it would be possible to write an hourly CI that does a check if the redirects land in other countries/continents and fetch mirrorlist to confirm 2021-04-27T12:27:33 the thing is - we should be seeing the same issue everywhere. I should be getting us mirrors - at least occasionally. 2021-04-27T12:27:55 not necessarily 2021-04-27T12:28:32 there are at least as many mirrors in Europe as there are for Asia and Americas combined 2021-04-27T12:28:41 it depends on the thresholds for these things 2021-04-27T12:32:02 when I get a mirrorlist for something TW, I have 3 mirrors same country, 43 other countries on the same continent, and 25 elsewhere. 2021-04-27T12:32:56 with a US address, I see 12 mirrors plus 59 elsewhere (other continents). 2021-04-27T12:34:52 * Eighth_Doctor shrugs 2021-04-27T12:35:10 Eighth_Doctor: yeah. 2021-04-27T13:41:14 lcp: nominally, 13, of which 2 provo, 3 kernel.org 2021-04-27T13:41:25 map link is useless 2021-04-27T13:41:30 The Google Maps Platform server rejected your request. You must use an API key to authenticate each request to Google Maps Platform APIs. For additional information, please refer to http://g.co/dev/maps-no-account 2021-04-27T13:42:07 closest mirror to me is probably more than 300 miles away 2021-04-27T13:42:50 I routinely get directed to Europe 2021-04-27T13:43:10 historically that is 2021-04-27T13:44:08 over 90 minutes to upgrade 1200 TW packages yesterday 2021-04-27T13:44:38 No Packman in USA 2021-04-27T13:46:26 a-865k: yes, we have an open ticket on the map link. 2021-04-27T13:47:15 a-865k: the kernel.org mirrors should be at least two different locations. 2021-04-27T14:21:04 12-13 mirrors ought to be plenty, but I guess the mb redirection uses a wider span for the round-robin selection. 2021-04-27T15:02:08 a-865k: fwiw, i have opened a ticket, progress#91836 2021-04-27T17:05:23 pjessen: :) 2021-04-27T22:01:40 pjessen: FYI: on mx1,2 there is a new file /etc/postfix/handling_special_recipients for the noreply@opensuse.org Email. Feel free to integrate it wherever your think it fits best :-) 2021-04-27T22:02:17 I switched OBS to use this Email for outgoing messages now. I hope this makes admin-auto more silent.