Description of problem: At intervals, bonobo-activation-server begins to eat all available CPU time. It becomes very difficult to open up desktop applications (such as gnome-terminal, to kill the process!). Version-Release number of selected component (if applicable): 2.14.0-1 How reproducible: Often Steps to Reproduce: 1. Log into desktop 2. Run some applications; it doesn't seem to hinge on which ones. 3. Observe CPU spikage, which continues unabated until I kill the b-a-s process. Additional info: FC5 + all current updates, including kernel-2.6.17-1.2174_FC5. At first I though this might be Evo related, but I wasn't even running Evo the last time this happened (I had run "killev" several minutes before). Once I do a "killall bonobo-activation-server" and the new instance starts, CPU usage returns to normal for some time. I ran strace earlier and unfortunately forgot to capture the log. I believe the error reported was that there were too many open files. I'm sorry I don't have more info; I am currently capturing an strace so I'll have a log if it recurs. I saw this issue also reported by Patrick Doyle on August 10 in the fedora-list archives.
Created attachment 134100 [details] gzipped strace output for b-a-s process I captured the bad behavior finally this morning in an strace. Unfortunately, it's about 750MB in size, but thankfully the reptition results in an attachment that's only about 3.4 MB gzipped. Just be warned when you uncompress it. I apologize for not snipping it more, but without more RAM I wasn't clear on a good way to edit it.
Thank you, that could be very useful! I'll have a look at it after the weekend.
I think I may have found the problem; I believe the mail-notification applet may be triggering this bug: http://bugzilla.gnome.org/show_bug.cgi?id=315650 I started using this applet at around the same time I did a batch of updates, so it wasn't immediately apparent to me what had changed. (Bad scientific method, sorry.) When I start the applet the results of $(ls -l `/sbin/pidof bonobo-activation-server` | wc -l) starts going up by about 4 per second. When I kill the mail-notification-applet, the leak stops. Looks like this was recently fixed in HEAD; any chance you'll pick this up for Fedora?
Sure, I can push a test update to see if it helps. When it hits 100% how many bonobo-activation-server processes do you see? Do you think this is a dupe of bug 189610?
Just the one, sorry.
The libbonobo you just rolled out to testing (2.14.0-1.fc5.1) appears to have taken care of this. Running a watch on the previous command noted in comment #3 indicates the leak's gone. Thanks Ray!
awesome, i'll push it live in a couple of days.
Hmm, I'm not sure this is related, maybe you can duplicate? Now Evo, xchat-gnome, and other applications won't quit correctly. They simply hang on close. I am *very* unknowledgeable about what libbonobo actually does, and it seems like the leaking file descriptor problem is still fixed... I'll play with it and open a new bug if I can't get any traction.
does the problem go away if you downgrade?
Please ignore comment #8. I think I was having an unrelated problem, either dbus or a hung esd. Everything working normally later yesterday and today. Thanks for your time and sorry for the bother.
Hi, We no longer support Fedora Core 5 and I am currently trying to get my open bug count down to a more manageable state. I'm going to close this bug as WONTFIX. If this issue is still a concern for you, would you mind trying to reproduce on a supported version of Fedora and reopening? (this is a mass message)