Description of problem: The problem is simple to describe. Sometimes (not every time) anaconda-yum will hange consumming approximately 100% of a VCPU (dual VCPUs defined). It will stop on various packages and usually different each time it occurs. This install is running on a qemu-kvm virtual system with 2 VCPUs, 2 GB memory, 8GB virtual disk and a created (mock/pungi) distribution DVD iso with 5/26/2014 rawhide as input. Version-Release number of selected component (if applicable): rawhide 5/26/2014 How reproducible: sometimes Steps to Reproduce: 1.define system 2.boot 3.select LXDE, regular partitions with ext4 Actual results: sometime it hangs Expected results: runs every time Additional info: Now if this was a network install I could maybe understand a hang occurring sometimes. But, this was a bloddy iso image on a host system disk. Since anaconda-yum runs as a separate process, I am not sure how to debug this or what to look for. Interesting, I inserted some import pdb/pdb.set_trace() in yumpayload.py and let anaconda-yum start but then stopped updating the display. When I did this anaconda-yum ran to completion every time.
I was incorrect about the trace. I put the trace in but it still hung.
Any tips on how to figure out what anaconda-yum is doing?
running strace -p <anaconda-yum pid> *seemed* to make it run for me if you start it early. strace after it hangs shows no activity at all, making it really hard to track this down. Something is going wrong inside yum/rpm/etc. but so far we don't have any good way to narrow it down.
I believe that this might take "instrumenting" anaconda-yum ... writing a whole bunch of debug info to some file so that you can get a better handle on what is going on. You are correct, this is going to be difficult. I am going to try going back to anaconda-21.36-1 (I have a DVD iso for that created on 5/20/14) and keep trying to install. If it is only recent, that might narrow the possibilities.
Bad news :( 1. Host is AMD 8150 with 8 cores 2. Three qemu-kvm virtual systems each has 2 VCPUs, 2GB memory and 8GB virtual disk. 3. A kickstart install for LXDE desktop and btrfs filesystem. 4. All three running at the same time. Two complete successfully. The third hangs and strace -p <pid> shows nothing. This is going to be a real son-of-a-xxxx
I forgot the most important part. The source was a distribution DVD iso created on 5/20/2014 with anaconda-2.36-1
Created attachment 900440 [details] strace -p <pid_of_anaconda-yum> Got it. Here is an strace where anaconda-yum is consuming 100% VCPU (dual VCPUs). The source was a DVD.iso created 5/29/2014 rawhide. I can almost (but not quite) create this on demand so if more info is needed, I can run this again. BTW, when it runs to completion, the strace file is over 100MB.
Well that's progress, sorta. Looks like it got stuck in a fdatasync on fd 36, but I can't see where that fd was opened. If you can reproduce this again (I may also try today) run a lsof -p <pid of anaconda-yum> to get a list of the open fds and just the tail end of the strace will probably be fine.
Unless something changed today, recent versions of rawhide do not work as far as doing an install is concerned. Going back to an installer created on the 29th of May seems to work so I may need to use that one. It is still occurring so I will see if I can get more data later today.
It isn't consistent. I was able to use the boot.iso from yesterday to do a couple of rawhide installations.
This bug appears to have been reported against 'rawhide' during the Fedora 22 development cycle. Changing version to '22'. More information and reason for this action is here: https://fedoraproject.org/wiki/Fedora_Program_Management/HouseKeeping/Fedora22
anaconda has switched to dnf by default and anaconda-yum is no longer going to be updated.
*** Bug 1146015 has been marked as a duplicate of this bug. ***