[Bug 260399] freebsd-update: Downloading patches often fails repeatedly

From: <bugzilla-noreply_at_freebsd.org>
Date: Mon, 20 Jun 2022 12:06:54 UTC

Julian Noble <julian@precisium.com.au> changed:

           What    |Removed                     |Added
                 CC|                            |julian@precisium.com.au

--- Comment #13 from Julian Noble <julian@precisium.com.au> ---
I think the most important bug here is the failure to restart utilizing the
already downloaded patches and files.

I'm usually running it on a low latency 1Gbps link in a big city in Australia -
but freebsd-update has been a big *slow* problem for me for a long time on
different systems and different network paths.
The additional latency and no local mirrors doesn't help - but all that would
be bearable if it just wouldn't restart doing work that has already been done
each time.

Really I don't think understanding exactly why a particular request failed or
exactly what version people are upgrading to or from is the issue. The networks
will never be perfect.

I've tried proxy/caching via nginx - but that's another layer of indirection
which makes it harder to debug (e.g it seems that 404s are common and normal -
but some can be cached - others shouldn't be?)
Simply doing a quick restart of nginx means any freebsd-updates pointed to it
need to restart from scratch... which suggests to me that the way this works
isn't really suitable for global distribution.

I'd love to try running my own freebsd-update server - but that seems a massive
job - and really I can see fewer and fewer technically inclined people even
getting to that level with the freebsd-update experience as it is.

Perhaps it's fine in North America with latencies of a few 10s of milliseconds
and everyone there barely cares about the odd restart from scratch?

I know freebsd-update involves a lot of cleverness in what it does - but I 
understand the frustrations of the OP. 
Starting a slow grind through 57000+ high latency http requests again is more
than a minor annoyance.  It's not just an hour or two either..  these update
attempts for a few machines can end up taking days of an admins time and thus
get put off. The only indication at the end is "failed"
In some cases - such as running via a wrapper such as iocage - I don't think
there's even a way to pass through the debug flag.

Is there a way to make freebsd-update resume from close to where it stopped?

You are receiving this mail because:
You are on the CC list for the bug.