[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: MD5 again

>>>>> "H" == Henrique de Moraes Holschuh <hmh@debian.org> writes:
H> Acquire::http::No-Cache=True
Yes, using since long ago.

H> There is also a second possibility: a single slow-to-update mirror in a
H> round-robin alias is enough to cause such failures.

Yes, every time people have looked at my reports that has been the case.
I think it is maybe because I look at the output of apt-get update more
than most people, who must be using various cron jobs, whereas I do it
by hand.

During a single apt-get update it might get the two[?] parts from two
different machines? Bad if true. Apt should not request a second IP
address during the same invocation, if that is what is going on.

H> I am not sure all tools are smart enough to retry with a different
H> IP from the round-robin alias when fetching indexes (if they're not,
H> it is a bug).

Or it will just further mask the lack of "first prepare all the parts of
dinner in the kitchen, before you present it to the guests in one atomic
write to disk"?

H> Meanwhile, set your sources.list to a *specific* mirror to ease your
H> debugging.  And there are probably mirrors closer to you network-wise than
H> ftp.us.debian.org, you might want to check that too.

If it really is caching that must be defeated, perhaps apt could add a
"add some random junk to the URL" feature, like
$RANDOM.www.mirror.debian.org:$RANDOM/....Index?$RANDOM&$RANDOM etc.
But I don't think that's my problem.
I'll choose rocky-mountain.csail.mit.edu for now, and write back if it
occurs again. Yes there is a .TW mirror but that just adds another layer
to debug.

Reply to: