On Wed, Jan 23, 2002 at 07:59:14PM -0800, Ian Clarke wrote: > On Thu, Jan 24, 2002 at 05:51:52AM +0200, Jukka Holappa wrote: > > I was just reading Ian's latest comments about datastore corruption > > from mail archives and how it might get corrupted when processes get > > killed unexpectly. > > > > Is it possible to install some kind of signal handlers in Fred so you can > > caught this signal (SIGTERM and perhaps others too) and shut down? This is > > how it's done in *nix world. > > Like most things in the world of Freenet, we discussed this so long ago I > can't remember the details, but I do recall that our conclusion was that > Java can't trap such things, at least, not in a cross-platform way. > > This was the origin of discussion around the addition of administration > commands to FCP which could do useful stuff like politely ask the node to > shut down cleanly - although the only result of this that I am aware of is > the addition of an optional administration password to the config file > (clearly you don't want just anyone logged into your computer shutting down > your node). > > Unless I hear a "Sure, it will be easy to guarantee that the datastore won't > get corrupted after an unexpected shutdown" from Tavin, perhaps it is time > to re-open this issue.
The core premise of the design of the new datastore is that it be completely impervious to termination of the application under any conditions which fall short of physically damaging the hard disk. That is not the difficult part. What's difficult is guaranteeing that buggy code doesn't corrupt the datastore while it is running.. -tc _______________________________________________ Devl mailing list Devl at freenetproject.org http://lists.freenetproject.org/mailman/listinfo/devl
