On Tue, Mar 09, 2004 at 11:14:45AM +0100, Markus Bertheau wrote: > On ???, 2004-03-08 at 10:15 -0700, Michael Stenner wrote: > > On Mon, Mar 08, 2004 at 11:05:21AM +0100, Markus Bertheau wrote: > > > Why deos it always fail? > > > > Perhaps you could give us a little more information. > > > > * yum version? > > * python version? > > yum-2.0.5.20040303-1 > python-2.3.3-2 > (This is an up-to-date Fedora Core Test1) I've already found the new version of python to create some issues with the url-fetching code, but I think that was a different bug. Just in case, I'd like you to run the following test script on a failing file. http://linux.duke.edu/~mstenner/test/dropped_keepalive_test.py I've now written this generically so that anybody can use it for any file. You'll need to tell python where to find keepalive and pass this script a currently-failing url. (the one from your original post is no longer on the server) For example, you should be able to do this: export PYTHONPATH=/usr/share/yum ./dropped_keepalive_test.py http://yourmirror.org/path/to/file.hdr Tell me what that does. If it produces any kind of negative result, please paste in the entire result. > > * you always show it failing on the same header: > > It always fails on the same header; see below. > > > - does it get headers.info? > > This is the only download that fails, that is, no message that hints to > the fact that another failure of any kind has occured before, is > printed. Yum will bitch mightily if something fails, so I'm inclined to say that if you didn't see anything, nothing else broke. > > - does it get other headers? > > This is the first header that yum tries to download. After a while the > download succeeded and it was apparently the only new header. On other > occasions it successfully downloads a number of headers, then > continuously (that is - upon retry) failing on the same header for an > extended period of time. The intermittent nature of this is consistent with keepalive issues. What was happening in the other case was that the server was dropping the connection after headers.info but before the first header (because of the time delay) and due to some changes in the core httplib module, an exception was being raised that keepalive didn't expect. Like I said above, though, this is a different exception and I couldn't recreate it on my system. > > - is this the first header? > > Yes. In other cases it was not the first. But when it fails it > continuously fails on the same header. I retry some times then and it > wouldn't work. When I try again later that day it works. > > If there is any other information you need, I will try my best to supply > it. If that script returns something interesting, let me know. Otherwise, you can also set DEBUG=1 in keepalive.py and rerun yum. That should provide a little more info (but not much). Then we'll take it from there. Oh, there's nothing else interesting in your setup, is there? Proxies, etc? -Michael -- Michael D. Stenner mstenner@xxxxxxxxxxxxxxx ECE Department, the University of Arizona 520-626-1619 1230 E. Speedway Blvd., Tucson, AZ 85721-0104 ECE 524G
