Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!swrinde!emory!stiatl!rsiatl!jgd From: jgd@rsiatl.UUCP (John G. De Armond) Newsgroups: comp.unix.i386 Subject: Re: sparing bad blocks Keywords: ISC mkpart Message-ID: <679@rsiatl.UUCP> Date: 27 Nov 89 11:43:20 GMT References: <5556@uhccux.uhcc.hawaii.edu> Reply-To: jgd@rsiatl.UUCP (John G. De Armond) Organization: Radiation Systems, Inc. (a thinktank, motorcycle, car and gun works facility) Lines: 73 In article <5556@uhccux.uhcc.hawaii.edu> bt455s10@uhccux.uhcc.hawaii.edu (Carl "Art" McIntosh) writes: >How does one map out bad sectors under ISC 2.02 ? >I have tried to map out a new bad sector reported >by a console message giving the absolute sector, >the drive and controller numbers. According to >the man page mkpart(1M) the -A option should do >the trick, however mkpart -A 266867 disk0 yields >the message V_REMOUNT failed and informs me a >reboot or remount or the device is required. >after rebooting the system, the /etc/partitions >file does not list the new bad sector, furthermore >mkpart -t vpa disk0 does not show sector 266867 >as having been remapped. Anyone know how this >supposedly simple procedure is done ? God, this is deja vu, having been working with large drive containing hundreds of bad sectors. :-) What you did worked, the problem is a matter of interpretation. The /etc/partitions file is a read-only prototype file for mkpart. It is NOT updated by mkpart. To view the Volume Table of Contents (VTOC), invoke mkpart with the command: mkpart -tvpa disk1 This makes mkpart print out statistics on the drive. I've found it invaluable to make a hardcopy of this table for when the volume gets wiped out. You don't have to rediscover the bad sectors all over again. You should be aware of a terrible problem with mkpart that makes it extremely dangerous. If /etc/partitions gets altered or corrupted and you invoke mkpart - even for a read-only status request as above - it will without your permission try to force the drive VTOC to agree with /etc/partitions. Usually this results in a trashed volume. I learned this the hard way when I moved a large drive off the root and installed a small, fast drive for the root partition. When I restored the root partition from tape, the previous /etc/partitions file got written out. After restoring the /usr and other partitions to the large drive, I used the -tvpa option to print out statistics on the new configuration. Voila! Trashed root partition. I've changed permissions on /etc/partitions to read-only, root AND keep a checksum handy to use before invoking mkpart. Next, a question for the interactive types. Why does the -V or -v option with an Adaptec 1542 SCSI controller not work? I can hear the drive retry several times on each bad sector but the program never reports one. When a bad sector is encountered in operation, I hear only one retry before the kernal reports a problem. I had to collect bad sector information by dd'ing from /dev/dsk/0s1 etc and logging the kernal messages. What a bitch! One more comment. If your drive/controller has a low level bad sector mapping facility, USE IT. The Interactive facility seems very fragile. I took the drive in question out of service and repeatedly assaulted it with the Adaptec firmware mapping program (HINT: reprogram the typmatic action on the keyboard for about 1 character per second and sit a large ball-pein hammer on the "Y" key :-). I let the program run until it would run for 24 hours without finding an additional bad sector. When I put the drive back in service, it was like a whole different drive. Where before it would grow a bad sector or 2 a day, it has been up for about 3 weeks with nary a peep. Hint to Adaptec: PLEASE give me a verify mode where I don't have to hit the damn 'Y' key for every bad sector. A batch mode would be fine. John -- John De Armond, WD4OQC | Manual? ... What manual ?!? Radiation Systems, Inc. Atlanta, GA | This is Unix, My son, You emory!rsiatl!jgd **I am the NRA** | just GOTTA Know!!!