I am running BOINC on several windows and Linux machines. My Linux machines (running Ubuntu) all give an error message of "no shared memory segment" fairly often when running a task. It appears that the task then re-starts from the beginning. This does not always happen, maybe 50% of the time. What can I do to fix this problem?

I am running BOINC 6.12.33 on Ubuntu 11.10 (kernel 3.0.0-12)

BOINC was installed from a repository.

I frequently see it from the Rosetta project, occasional from Einstein and WCG

There doesn't seem to be an error number. The message just lists out the name of the task followed by "no shared memory segment". the next message line then states "task xxxxxxxx exited with zero status but no finished file.

Not sure of what you mean by a link to each computer...

SETI and SIMAP work units have never shown this error.

Does not help much to post here about task failures at projects. Please connect over at the WCG forums and copy/paste the My Grid > Result Status > Status link (Error/Invalid) and post content. Smells like heartbeat issue, and with Linux 11.10 observing wild running dbus-daemon and zeitgeist processes and file indexers, taking serious juice, each. Going into terminal (Ctrl-Alt-F1 e.g.) and exiting the GUI by entering e.g. sudo service lightdm stop (or instead of lightdm gdm or kdm or whatever runs the signin screen) and all these processes stop and BOINC gets 99.9% of all cycles. Check with "top -u boinc" command. This is how I run Linux most of the time, it co-acting as a file / backup server.

--//--

edit: appeared Zeitgeist was driving the dbus-deamon hi CPU usage. Disable that and got 75% back of 1 core while in GUI mode.