Obviously this is a highly transient issue, but right now - Tue Dec 8 19:01:12 UTC 2015 - I caught mirrormanager definitely offering old checksums for repomd.xml. If you hit https://mirrors.fedoraproject.org/metalink?repo=rawhide&arch=x86_64 right at this time, it returns this:
<file name="repomd.xml"> <mm0:timestamp>1449476392</mm0:timestamp> <size>4310</size> <verification> <hash type="md5">a029eb6ec46181666e9cbf663cde1f42</hash> <hash type="sha1">bbe3acef1f20215599da7dcd2ebdc0b93049cdf2</hash> <hash type="sha256">f4e696fe5aad7e063cb1acae3adfae0159274414ae18777eb1e0dcb71fbd9ad7</hash> <hash type="sha512">55d0734a66ba8e5b1419d4f3b4b6e6f6fcd118890bdbb78f944843b6afb0adc34455515b6499374bfea6f73118254d30a5ec52eb2c773e6e77a5191dc5ef2014</hash> </verification> <mm0:alternates> <mm0:alternate> <mm0:timestamp>1449389233</mm0:timestamp> <size>4310</size> <verification> <hash type="md5">1b49c64edfccc2edc89f3455810508fe</hash> <hash type="sha1">2e8c65b451be3e09b1f16586b86b557d24096831</hash> <hash type="sha256">4b0c2c24a6088a3c97834997037857bfab77a75db4ebac3d070d5c94b786d0da</hash> <hash type="sha512">0be5bb77735989f77aa0d2b5e5a6ab85332952c66210dc1560306031643a6364f0bae4bf00b0e0ae8983f758b88d2b1c6bf778a6f43416b59d368233e33ab990</hash>
Those timestamps are old - the first is Mon, 07 Dec 2015 08:19:52 GMT , the second is Sun, 06 Dec 2015 08:07:13 GMT - and indeed the metadata they refer to is old. As I write this, a repomd.xml with the first checksum was still present in http://www.nic.funet.fi/pub/mirrors/fedora.redhat.com/pub/fedora/linux/development/rawhide/x86_64/os/repodata/ , clearly date-stamped Dec 07, and a repomd.xml with the second checksum was still present in http://mirror.csclub.uwaterloo.ca/fedora/linux/development/rawhide/x86_64/os/repodata/ , clearly date-stamped Dec 06. But the actual current repomd.xml , as found at https://dl.fedoraproject.org/pub/fedora/linux/development/rawhide/x86_64/os/repodata/ , is date-stamped Dec 08 and has a sha512sum of 0c3a8664f3c27646e5e68c73ad3d0cd6856f71cdc517e5a622ef2c45459204dab9c45aeb4c05740401ccafe4a658f5463263b1d0cb32359ca73c69d97def44cb , which is not present in the metalink at all.
Consequently, when I just ran a Rawhide net install, it rejected my up-to-date local mirror and a bunch of other up-to-date mirrors and wound up using the two-days-out-of-date uwaterloo mirror.
Note: the precise date stamp on the 2015-12-08 repomd.xml is 2015-12-08 08:09 (that should be UTC, I believe, as dl.fp.o has UTC set as its time zone) and I was testing at 19:01:12 UTC as mentioned above, so this isn't some very tiny window, the new repomd.xml has been up for 11 hours.
I think this conversation was about this bug:
<nirik> oh, I bet it's related to rawhide composes not sending fedmsgs... so it only runs daily instead of fedmsg triggered. <threebean> ah, right. <threebean> that should hopefully be fixed now. <nirik> well, fixed with tomorrows rawhide right? * threebean nods
Still happening, the timestamp on metalink right now is Fri, 11 Dec 2015 07:49:00 GMT, dl.fp.o has a repomd.xml timestamped 2015-12-12 07:54 .
yep. Still having no luck getting rawhide compose fedmsgs working again. :(
We just pushed another change this morning that hopefully will work. I guess we will see tomorrow.
Finally we are back in business.
There were rawhide compose fedmsgs today and mm seems to have picked up correctly on it after it finished. ;)
Please let us know if you still see any issues with it...
yeah, it seems good! Thanks.
Login to comment on this ticket.