New AccessGrid-3.1 RPM has been released for Fedora 9 & 10 and RHEL/CentOS 5 which fixes the dbus/Avahi intermittent crashing issue when starting up the VenueClient.
It took a little longer than expected to release as I needed to work out some issues killing the /usr/bin/avahi-publish-service processes that get forked from the VenueClient. Doug From: ag-tech-boun...@lists.mcs.anl.gov [mailto:ag-tech-boun...@lists.mcs.anl.gov] On Behalf Of Douglas Kosovic Sent: Wednesday, 4 March 2009 5:36 PM To: Khanh ly Cc: ag-t...@lists.mcs.anl.gov Subject: Re: [AG-TECH] Inconsistent start up error - FC 10 and AGTK 3.x Hi Khanh, The main issue with the workaround is that if you want to do any node services configuration (i.e. Tools menu -> "Configure node services..."), it won't automatically discover the service URLs. The new RPM is just going to have a modified version of ServiceDiscovery.py which invokes /usr/bin/avahi-publish-service instead of using Avahi Python routines to publish services. Unless you will be doing any node services configuration, you won't notice a difference with the new RPM. Doug From: Khanh ly [mailto:khan...@wasp.uwa.edu.au] Sent: Wednesday, 4 March 2009 5:00 PM To: Douglas Kosovic Cc: ag-t...@lists.mcs.anl.gov; khan...@wasp.uwa.edu.au Subject: Re: [AG-TECH] Inconsistent start up error - FC 10 and AGTK 3.x Hi Doug, Thanks for your workaround solution. It worked and save me some headache in troubleshooting AG :-). Yes. Our CPU is a quad core Intel E5420. We will do more intensive test this week. However, I wonder should I use your workaround or update to the promised new AG RPM later. I also thanks Andrew for a troubleshooting of the multicast network suggestion which we are likely to use if we encounter other problems. Regards Khanh Hi Khanh, We set up a single user node with the following configuration: FC 10 and AGTK 3.x, Multicast network We have inconsistent and random errors whenever we start AG. However, after two or three times failed to start then we can start AG client successfully. We can enter Asia Pacific venue, and the audio and video seem OK for more than two hours. The intermittent startup Python segmentation issue is related to a 'dbus' multi-threading bug, some machines are more prone to it than others (e.g quad core CPUs seem to be more prone to it than dual cores). I will be releasing a new AccessGrid RPM which uses an external single threaded program for the Avahi/dbus stuff and should be more stable. If you want a workaround now, disable Avahi/dbus support in the VenueClient by editing: /usr/lib/python2.5/site-packages/AccessGrid3/AccessGrid/ServiceDiscovery.py and change: import avahi haveAvahi = True to: #import avahi #haveAvahi = True haveAvahi = False Cheers, Doug
