Bugzilla – Bug 4922
middleware drivers are not reloaded when they crash
Last modified: 2007-06-21 07:08:44
You need to
before you can comment on or make changes to this bug.
When a driver hangs (or for that matter when the process finshes) GridWay does
not reloads it, so the communication stream between GridWay and the driver is
no longer valid. This makes GridWay to continuously output "Error reading MAD
message" in the gwd.log, so virtually freezing the daemon.
This behavior has some side effects when recovering GridWay, if the proxy of a
user is no longer available the driver will not load and the job will not be
recovered, hanging the recovering process.
Moreover, in some situations the ability to reload middleware drivers (for
example if the driver hangs) will come in handy, so the daemon has not to be
restarted in such case.