| [Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] |
> -----Original Message----- > From: Greg Freemyer [mailto:freemyer-ml@NorcrossGroup.com] > Sent: Wednesday, September 17, 2003 12:26 PM > To: opengfs-users@lists.sourceforge.net > Subject: Re: [ogfs-users]OpenGFS viability > > > Brian, > > The current lock server implementation in ogfs is a single point of > failure. > > I'm not sure what the fall-out is if the node running it dies is, but > that is likely the biggest problem from a production perspective. Ben > has a very detailed write-up about how the locking works, but I don't > know if he addresses the failure case or not. No, the failure case is not covered in the docs. I remember some talk about using Linux-HA as a failover mechanism for the lock (storage) server (the daemon program is called memexpd), but I don't think anyone has investigated much. IIRC, the whole cluster panics if the central lock storage server goes down. Not good. We are working on getting OpenDLM implemented as an alternative locking system. More general comments on production-worthiness: Recent testing shows that the filesystem is reasonably stable, but beware that the filesystem will panic if it gets full. Also, performance is not nearly what the shared architecture should be capable of delivering. We need to do some work to find out why. If the number of folks who could contribute to the project were as high as the number of folks who want to use it, we could get things happening faster. But it's a tough nut to crack ... a very complex system. I've been hoping that documentation of internals would lower the entry barrier, but there's still a lot to learn. -- Ben -- Opinions are mine, not Intel's ------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf _______________________________________________ Opengfs-users mailing list Opengfs-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/opengfs-users