Package Versions Are 404 With Many Mirrors

vindicator · 2026-04-24 03:33:40

Incorrectly posted at: https://bbs.archlinux.org/viewtopic.php … 5#p2295595

I had just commented on the topic in IRC.
I just didn't know if this was the first time ever that I would've encountered a file so freshly built that didn't yet sync with the providing mirror I chose.

I was updating, when I encountered a 404, even though I used `-Sy`, but even with `-Syy`, I was still getting a 404 for `jq-1.8.1-2-x86_64` package at the kernel.org mirror (also `libqaccessibilityclient-qt6-0.6.0-3-x86_64` when testing more).
I started uncommenting servers I thought would be reasonable, but more 404's for the jq package kept adding up.
I then used `--debug` and sed uncommented every server, and there were like 21 servers that 404'd before getting to `mirror.xeonbd.com` which had it I guess, but errored with: `Exceeded the maximum allowed file size (16384) with 16384 bytes` for the sig file. Same with the `au.mirrors.cicku.me` mirror.
The jq and accessibility packages were considered corrupt so were deleted and re-downloaded.

I kept those unproblematic mirrors uncommented and got all of the packages now, and it's now starting the unpacking step.

As an aside from the other post, I wasn't aware of "archives" (or maybe it's just been many years since the topic came up). It's great to know about, and something I expect to look into more.

seth · 2026-04-24 07:58:53

https://archlinux.org/mirrors/status/
mirrors.kernel.org recently has some problems, https://bbs.archlinux.org/viewtopic.php?id=313214 (AI DDOS?) - not sure what might be wrong w/ the others, did you -Syyu on one of them (in case you got a bogus database from kernel.org?)

Then see eg. https://archlinux.org/packages/extra/any/reflector/

vindicator · 2026-04-24 19:54:35

Yup, I `-yy`'d as well: "but even with `-Syy`, I was still getting a 404 for `jq-1.8.1-2-x86_64`" (insinuating that I force a fresh pull of the catalogs, along with updating).

And it was just that jq and the accessibility packages. The others before that pulled just fine from the kernel mirror. That just made me think I had possibly stumbled on a very rare case (if possible) of a catalog index being updated first before the actual package was synced. Does such a race-condition exist in the mirror syncing process?

Yeah, I believe I had previously posted about the kernel mirror missing from the packaged list last month (heh, just noticed you linked my post in the other post you linked here). I've only been seeing ipv6 being the issue on the status page (not an issue for me), and I believe this latest mirror package had the kernel mirror included this time.
I was surprised to see the kernel mirror so poorly positioned in "Successfully Syncing Mirrors". I've been using it forever. I could see that "race" case being more probable now (though I'd expect a mirror would maintain the previous sync of packages and catalog until the latest sync completes).
I went ahead and added a 100% one from that list (nice tracking/ranking service). Perhaps having a current dir of everything synced, and an updating dir. Then when the sync is complete, just rename the updating dir to current so it's seamless.

seth · 2026-04-24 20:13:37

Does such a race-condition exist in the mirror syncing process?

Yes, but that would be a one-off and a -Syu attempt some minutes later typically just work - except that mirrors.kernel.org seems to be under undue pressure recently.

Arch Linux

#1 2026-04-24 03:33:40

Package Versions Are 404 With Many Mirrors

#2 2026-04-24 07:58:53

Re: Package Versions Are 404 With Many Mirrors

#3 2026-04-24 19:54:35

Re: Package Versions Are 404 With Many Mirrors

#4 2026-04-24 20:13:37

Re: Package Versions Are 404 With Many Mirrors

Board footer