This is part of The Pile, a partial archive of some open source mailing lists and newsgroups.
Date: Mon, 28 Jan 2002 14:29:16 -0800 From: Jeffrey Siegal <jbs@quiotix.com> To: Darlene Wallach <wallachd@earthlink.net> Subject: Re: [svlug] question re partitioning system Darlene Wallach wrote: > Jeffrey Siegal wrote: > > > Walter Reed wrote: > >>Hmm. My take on this is different. I'd rather take the > >>extremely rare chance of corruption with ext3 than the > >>much higher risk of data loss (and the slow reboots) due > >>to abrupt shutdowns on ext2. > > That assumes you have abrupt shutdowns. If you use even > > a very small UPS (with some sort of mechanism for > > detecting power failures, which could be the UPS itself, > > or just an old modem *not* plugged into the UPS), you > > should never have them. The only reason would be a > > kernel crash, which is very rare. I've not had an > > abrupt shutdown in several years of running Linux > > servers. > I'm taking a Linux Device Drivers class. Since I will be > trying to write Linux device drivers, doesn't that > increase my chances of abrupt shutdowns? Absolutely. Depending on what kind of drivers you're working on, you might want to take a look at UML (user mode Linux). UML lets you run a Linux kernel as an ordinary program, withing another Linux system. If the UML kernel crashes, your actual kernel just keeps going (it is just like a regular program crashing). One thing that I've done for device driver development in the past (before UML) was to put all of your dynamic files on another system using NFS, and just have static files (or no files at all, with some sort of net booting mechanism) on the system you're testing the drivers on.