Description of problem: TFTP is very slow in some situations with high latencies no matter the bandwidth. It takes up to 10 minutes to boot kickstart's vmlinuz and/or initrd.img Version-Release number of selected component (if applicable): Satellite 6 versions How reproducible: Each time Steps to Reproduce: 1. Register a host 2. Use a MacroLan/WAN connection with more than 30ms of latency, but good enough bandwidth (10Mbps) 3. Build a host Actual results: vmlinuz and initrd.img takes around 12 minutes to be transfered Expected results: Having vmlinuz and initrd.img transfered in better times according to connection bandwidth (a couple of minutes). Additional info: As a workaround we have tweaked block size and timeouts to retransmit in tftp server, but the speed is still low. As a solution, we have made next changes in Satellite 6 and I would like to ensure if it's supported or not from Red Hat: - used a modified clone of "Kickstart default PXELinux" template to use <%= @host.url_for_boot(:initrd) -%> instead of <%= @initrd %>, and so for vmlinuz. - copied /usr/share/syslinux/gpxelinuxk.0 to /var/lib/tftpboot - modified dhcp parameters to provide gpxelinuxk.0 as filename. As a result: - a client boots from DHCP and receives gpxelinuxk.0 from the DHCP server as filename - it gets the file from Satellite using TFTP - it gets the boot file from Satellite using TFTP - it gets vmlinuz and/or initrd.img from Satellite using HTTP - the total deployment time reduces from 20 minutes (being 10 only to transfer vmlinuz/initrd.img) to 10 minutes.
Since this issue was entered in Red Hat Bugzilla, the release flag has been set to ? to ensure that it is properly evaluated for this release.
This usage case is already supported, see http://projects.theforeman.org/projects/foreman/wiki/Fetch_boot_files_via_http_instead_of_TFTP for upstream instructions.
I am closing this bug, there is nothing to do in the code base. This is only documentation effort. I filed new bug which is here to track documentation of this feature and how to configure it. https://bugzilla.redhat.com/show_bug.cgi?id=1308896 *** This bug has been marked as a duplicate of bug 1308896 ***