It's possible to do a variation of what you describe by adjusting environemnt variables that specify the length of the timeout and the maximum number. I don't think this behavior should be the default, but by adjusting those values, you could get what you're after.
I'm hesitant to spend a lot of time on this particular aspect of the code. Once we have mirroring support, we can favor mirrors that are faster and have fewer errors. It seems more worthwhile to work on solving that problem than to fine tune the timeout handling now -- it's going to change as we add mirror support anyway. -j On Tue, Jun 10, 2008 at 05:36:11PM -0700, Brock Pytlik wrote: > [EMAIL PROTECTED] wrote: > > Keep in mind that this is number of retries per download operation. > > We'd retry 3 times to get a manifest, just as we'd retry three times to > > get ~1mb of file data. Retrying 20-30 times for files would mean that > > you could end up waiting up to 15 minutes for a chunk, in the worst > > case. That hardly seems acceptable to me. > > > > -j > > > I agree it's not acceptable... on the other hand, if I start my > image-update just before I go home for the night, I'd much rather it > just kept trying to get the file while I'm sleeping than just giving up. > >From that perspective, there's no difference between it taking 8 mins or > 8 hours as long as it's finished when I get back the next morning (for > example). That was the situation I was envisioning when I talked about > setting the retry level to 20 or 30. Of course, from that perspective, I > might want to use the same level of redundancy for the manifest files. > > I think the idea I was thinking of was roughly using manifest retrieval > as a canary to determine whether getting the files had a good chance of > succeeding or not (from a user perspective). Maybe this isn't a viable > approach. > > Brock > _______________________________________________ > pkg-discuss mailing list > [email protected] > http://mail.opensolaris.org/mailman/listinfo/pkg-discuss _______________________________________________ pkg-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
