Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!uunet!ogicse!littlei!gandalf!andyc From: andyc@bucky.intel.com (Andy Crump) Newsgroups: comp.unix.sysv386 Subject: Re: SVR4 ufs file system reliability Message-ID: Date: 14 Dec 90 09:18:05 GMT References: <7454@suns302.cel.co.uk> <1630@svin02.info.win.tue.nl> Sender: news@littlei.UUCP Organization: Intel Corporation, Hillsboro, Oregon Lines: 32 In-reply-to: tim@delluk.uucp's message of 11 Dec 90 08:59:44 GMT >I have experimented with sVr4.0 version 2.0 and can only say that the >ufs file system is horribly unreliable. Shortly after reading about >100 mbytes from a tape (but well after the time the automatic update >program waits to write the stuff out to the disk) I got a panic, >and fsck went on and on complaining about my file systems... >I went back to sVr3.2. >Apparently the sVr4.0 ufs file system doesn't get sync-ed properly >by the automatic syncing deamon... In version 2.0 there are small windows in the VM subsystem that will cause a panic and corrupt the ufs filesystem. Generally these have not been a problem. I had been running version 2.0 with only ufs filesystems for 3 months on my workstation and had no problems. Problems with version 2.0 and ufs that I know will hurt you is if you have a filesytem with greater than 64k inodes. This will definitely hose you. Generally, I was able to create ufs filessystems in the 360meg range safely. To my knowledge these problems are all being addressed in the next version of SVR4 which should be alot more stable. -- -- Andy Crump ...!tektronix!reed!littlei!andyc | andyc@littlei.intel.com ...!uunet!littlei!andyc | andyc@littlei.uu.net Disclaimer: Any opinions expressed here are my own and not representive of Intel Corportation.