On Wed, Jan 15, 2003 at 03:51:38PM -0500, Ryan Babchishin wrote: > I was being sarcastic. It was a silly suggestion in the first place.
<whew> > There are no real low usage hours in my environment. There are lower > usage hours. Is this perhaps what I see at www.epalscorp.com? If so, maybe an outage could be disguised as some special feature, kind of a recess, goofing off approach. Come up with a game or contest or something that can be managed by a substitute machine. Maybe the "recess" can be a regularly weekend feature that runs for a few exclusive hours on Saturday...? I was reading something a few months ago on people striving for 5-nines reliability. I didn't like what I read but it did make two points worth thinking about (if I am remembering correctly): - .99999 is damn little downtime, so little that .9999 (only four 9s there that time) is still not so great. 5 minutes over the course of a year. - Scheduled downtime is valuable. One doesn't have to bring the system down on the schedule, but any necessary outages should be then. And *scheduled* downtime doesn't have to enter into the tallying of 9s. The annoying part was the article denying the need for extreme reliability. But maybe some downtime could be scheduled, it might be very useful. > but at the moment it's realy only a minor filesystem error that our > custom software doesn't deal well with. If things are working mostly, maybe it is possible to migrate to a collection of smaller file systems that fsck faster. A journaling file system might also help, but something tells me whatever gremlin got you this time would have sneaked past ext3 or jfs, etc. Good luck. Let us know how it turns out--if you can, -kb -- redhat-list mailing list unsubscribe mailto:[EMAIL PROTECTED]?subject=unsubscribe https://listman.redhat.com/mailman/listinfo/redhat-list
