todd@ivucsb.sba.ca.us (Todd Day) (08/12/90)
I've been thinking about getting the John Milton disk mod and getting two hard drives. I'd like to get about a 100MB (formatted) drive with under 20ms access time. Then, I want to retire my current 67MB(formatted) drive and use it for news only. It has served me well these last three years, but I think it's time to move on (plus, it is fragmented as hell). Anyway, what I am wondering, is: What is the least painful way to copy my entire 67MB drive to the 100MB drive and then make the 100MB drive my new root drive? Keep in mind that I want to somehow unfragment the files in the process, so I think using dd is out of the questiong. Also, any comments or suggestions concerning my choice of new drive would be appreciated... -- Todd Day | todd@ivucsb.sba.ca.us | ucsbcsl!ivucsb!todd "I believed what I was told, I thought it was a good life, I thought I was happy. Then I found something that changed it all..." --- Anonymous, 2112
mdapoz@hybrid.UUCP (Mark Dapoz) (08/12/90)
In article <1990Aug11.184541.273@ivucsb.sba.ca.us> todd@ivucsb.sba.ca.us (Todd Day) writes: >Anyway, what I am wondering, is: What is the least painful way >to copy my entire 67MB drive to the 100MB drive and then make >the 100MB drive my new root drive? Keep in mind that I want to >somehow unfragment the files in the process, so I think using >dd is out of the questiong. I went through this dillema about 3 times as I kept changing hard drive configurations. The easiest way I found to do it was to add the 100 meg drive as a second hard drive, format it, allocate the boot and swap partitions and then mount the user parition(s) (ie. /dev/fp012 onward) as something like /mnt on the running system. To copy the files to it you can use a find/cpio pipeline to do it. This will defragemnt the drive while copying and will also be very quick (well, relatively quick, the drives are slow). I've tried using a find/cpio pipeline when booted off of a floppy but it's very slow because the kernel i/o device for the pipe is the floppy instead of a hard drive. I don't recommend doing the copy using the floppy kernel unless you have a few days to spare. Once you've copied all the files, you can then switch the two drives and boot off the 100 meg one. You may have to use iv to create the boot partition, I don't think the diag disk does this for you. Hope this helps. -- Managing a software development team | Mark Dapoz is a lot like being on the psychiatric | mdapoz%hybrid@cs.toronto.edu ward. -Mitch Kapor, San Jose Mercury | mdapoz@torvm3.iinus1.ibm.com
lenny@icus.ICUS.COM (Lenny Tropiano) (08/12/90)
In article <1990Aug11.184541.273@ivucsb.sba.ca.us> todd@ivucsb.sba.ca.us (Todd Day) writes: |>I've been thinking about getting the John Milton disk mod |>and getting two hard drives. I'd like to get about a 100MB |>(formatted) drive with under 20ms access time. Then, I want [...] |>Anyway, what I am wondering, is: What is the least painful way |>to copy my entire 67MB drive to the 100MB drive and then make |>the 100MB drive my new root drive? Keep in mind that I want to |>somehow unfragment the files in the process, so I think using |>dd is out of the questiong. Well since we don't have a nice utility called dcopy(1M) on the 3B1, we'll have to settle for the next best thing to copy entire filesystems ... (BTW: For those who wonder, dcopy is a utility to copy one filesystem to another, optimizing seek performance based on last accessed files, it also will group all the blocks so they are contiguous. I was told this week by John McMillian (jcm@pegasus.att.com) that there is something even better called fscopy? that takes into account size as well as access time. Of course we'll never see something like that for the 3B1 from AT&T ... volcopy(1M) [also not showing up in our UNIX command subset] copies byte for byte, bit for bit ... so no defragmentation would occur) Anyhow, back to what I was saying, the next best thing to use is cpio, yes, cpio. # find / -print | sort > /tmp/FILES (I like sorting my files..) # vi /tmp/FILES (Remove anything you don't want copied) # mount /dev/fp012 /mnt (Mount that new disk slice of yours... Note root will HAVE to be slice 2, since slice 1 is swap) # cat /tmp/FILES | cpio -pduvm /mnt [...] Sit back and wait as the filesystems get copied over. # /etc/ldrcpy /dev/rfp000 /dev/rfp010 (This will put a boot loader on the second disk ... copied from the first disk) Now your set, swap the drives (remembering to swap the drive selects as well) and you'll be set, with a nice and fast, unfragmented root partition. |>Also, any comments or suggestions concerning my choice of new |>drive would be appreciated... |> 100MB formatted is an odd amount, probably have to go for something like a Maxtor XT-1140, which is ~113MB formatted, of course you could push the XT-1140 to the full 1224 cylinders and get ~151MB... The Maxtor drive seek times aren't anything terrific though... Let us know what you decide for a drive ... and how it works out. -Lenny -- | Lenny Tropiano ICUS Software Systems lenny@icus.ICUS.COM | | {ames,pacbell,decuac,sbcs,hombre,rayssd}!icus!lenny attmail!icus!lenny | +------ ICUS Software Systems -- PO Box 1; Islip Terrace, NY 11752 ------+
bdb@becker.UUCP (Bruce Becker) (08/12/90)
In article <1990Aug11.184541.273@ivucsb.sba.ca.us> todd@ivucsb.sba.ca.us (Todd Day) writes: >[...] >Anyway, what I am wondering, is: What is the least painful way >to copy my entire 67MB drive to the 100MB drive and then make >the 100MB drive my new root drive? Keep in mind that I want to >somehow unfragment the files in the process, so I think using >dd is out of the questiong. "dd" won't work because it copies _everything_, including the information about the amount of disk space used in the file system. (I suppose you could make a partition of the exact number of blocks as the original, but...) Anyway, what you want is find /mnt -depth -print | cpio -pdmu /mnt2 which would be done by making up a floppy file system with at least find and cpio on it. You can make a copy of disk 3 of the O/S foundation set of the installation floppys, and modify its etc/profile so as not to take you thru the installation procedure. Once you've installed the mod for 2 drives, boot the Floppy Boot disk 2 of the foundation set, and then use your newly-created floppy file system disk when prompted for disk 3. Then proceed to mount the old drive on /mnt; then mkdir /mnt2 and mount the new drive on it (which you've already formatted & made a file system on). Then you can issue the find/cpio sequence shown above. Since you're writing to a new file system, it will of course be unfragmented... Cheers, -- ,u, Bruce Becker Toronto, Ontario a /i/ Internet: bdb@becker.UUCP, bruce@gpu.utcs.toronto.edu `\o\-e UUCP: ...!uunet!mnetor!becker!bdb _< /_ "I still have my phil-os-o-phy" - Meredith Monk
bri@tsnews.Convergent.COM (Brian Rice) (08/13/90)
Um... I think I would go with Lenny's version (No offense intended to Mr. Becker) merely because of the sort/discard option he suggests. However, I don't recall seeing the word 'backup' in either of the follow-ups! I am currently running the Maxtor 2190's (Thanks Thad) and am also in agreement with Lenny on this: the seek time is not very good. All that space on the other hand...
jbm@celebr.uucp (John B. Milton) (08/14/90)
In article <1990Aug11.184541.273@ivucsb.sba.ca.us> todd@ivucsb.sba.ca.us (Todd Day) writes: >I've been thinking about getting the John Milton disk mod >and getting two hard drives. I'd like to get about a 100MB >(formatted) drive with under 20ms access time. Then, I want >to retire my current 67MB(formatted) drive and use it for >news only. It has served me well these last three years, but >I think it's time to move on (plus, it is fragmented as hell). > >Anyway, what I am wondering, is: What is the least painful way >to copy my entire 67MB drive to the 100MB drive and then make >the 100MB drive my new root drive? Keep in mind that I want to >somehow unfragment the files in the process, so I think using >dd is out of the questiong. > >Also, any comments or suggestions concerning my choice of new >drive would be appreciated... Setup the hardware so that the new 100 meg disk (will be root disk) is the second hard drive. Format it from the diag disk: use "s4test" to get to expert mode, then use the commands "2,2" to format and "2,23" to check. Use the new diag if you want to partition it. Reboot with the original 67M as (still) the first disk. Make the new disk bootable: # /etc/ldrcpy /dev/rfp000 /dev/rfp010 # mkfs /dev/rfp012 # fsck /dev/rfp012 Probably the best way to transfer the files is to generate a list of files first, making sure the target disk is NOT MOUNTED. Do everything with root permission. # cd / # find . -type d -print > /tmp/dlist # find . ! -type d -print > /tmp/dlist Edit the lists into the order you want things to be created. Put things like /bin/sh and /bin/ls first. Put programs that are not used by shell scripts (or not used at all) further down. Put news last on both the directory list and the file list. Delete files in /tmp, /usr/tmp from the list. One could agonize over the correct ordering of the list forever. # mount /dev/fp012 /mnt # cd / # </tmp/dlist cpio -pdm /mnt # </tmp/list cpio -pdm /mnt CHECK EVERYTHING ON /mnt TO MAKE SURE IT MADE IT!! Shutdown Set the select on the 67M for drive 2 and the select on the 100M to drive 1. Reboot and MAKE DAMN SURE everything is ok. Reboot to the diag disk. Reformat and verify the 67M (WHICH IS NOW THE SECOND DRIVE!!!) with expert 2,2 and 2,23. When it asks you about the size of the swap partition, answer 0. Our OS can only support one swap area. This will give you another 5M for news. Reboot to UNIX. # iv -t /dev/rfp010 Note the number of logical (1k) blocks of partition 2. # mkfs /dev/rfp012 <1k blocks>:<inodes> <1k blocks> is the number from the iv command. <inodes> can be as big as 65500. # fsck /dev/rfp012 # mv /usr/spool/news /usr/spool/onews # mkdir /usr/spool/news # new mount point # chmod 775 /usr/spool/news # fix the mount point # chown news /usr/spool/news # chgrp news /usr/spool/news # mount /dev/fp012 /usr/spool/news # mount the second disk # chmod 775 /usr/spool/news # fix the root dirs on the 2nd disk # chown news /usr/spool/news # chgrp news /usr/spool/news Setup mountable and checklist: # cat >>/etc/mountable mount /dev/fp012 /usr/spool/news ^D # cat >>/etc/checklist /dev/rfp012 ^D # cd /usr/spool/onews # find . -depth -print|cpio -pdm ../news Check to make sure news is really there in the /usr/spool/news tree. # rm -rf /usr/spool/onews Reboot to make sure everything gets FSCKed and MOUNTed properly. Everything will be, of course, unfragmented. John -- John Bly Milton IV, jbm@uncle.UUCP, n8emr!uncle!jbm@osu-cis.cis.ohio-state.edu (614) h:252-8544, w:469-1990; N8KSN, AMPR: 44.70.0.52; Don't FLAME, inform!
lenny@icus.ICUS.COM (Lenny Tropiano) (08/14/90)
In article <632@tsnews.Convergent.COM> bri@tsnews.Convergent.COM (Brian Rice) writes: |> |>Um... I think I would go with Lenny's version (No offense intended to Mr. |>Becker) merely because of the sort/discard option he suggests. However, |>I don't recall seeing the word 'backup' in either of the follow-ups! |>I am currently running the Maxtor 2190's (Thanks Thad) and am also in |>agreement with Lenny on this: the seek time is not very good. All that |>space on the other hand... You're right, but I think those who are thinking about this procedure know that you copying the root filesystem from one drive to another, therefore by the very nature of that operation you have a backup. :-) I guess the disclaimer should be ... *************************************************************************** DON'T FORMAT, MKFS, OR DESTROY YOUR ORIGINAL ROOT FILESYSTEM UNTIL YOU VERIFY THE NEW ONE ACTUALLY WORKS AND EVERYTHING YOU WANT IS THERE! *************************************************************************** Yes, folks even the best of us goof once and a while. I remember helping a good friend of mine swap some HD's around, while I was at her console (su'd as root) I decided to re-mkfs the slice 2 on the second drive (for those that don't know it's /dev/rfp012). Well as things go in those intense 'hacking' sessions... a slip up of one little number and POOF! Well I hate to admit it (and now as she is probably reading this, she's rolling in her seat ...) I typed: # mkfs /dev/rfp002 Tick... tick... tick. Aborting the procedure prematurely, but too late to save anything ... And feeling very embarrased, as well as very much responsible -- I non-chelantly [sic] said, "Do you have a recent backup of your root filesystem?" Of course I was handled the 90+ diskettes. [I guess that's better than most UNIXpc'ers] Two hours and then some later, things were back to some semblance ... The moral of the story is ... BACKUP, IT'S WORTH THE EFFORT! -Lenny -- | Lenny Tropiano ICUS Software Systems lenny@icus.ICUS.COM | | {ames,pacbell,decuac,sbcs,hombre,rayssd}!icus!lenny attmail!icus!lenny | +------ ICUS Software Systems -- PO Box 1; Islip Terrace, NY 11752 ------+
upl@gumby.cs.wisc.edu (Undergrad Projects Lab) (08/15/90)
In article <632@tsnews.Convergent.COM> bri@tsnews.Convergent.COM (Brian Rice) writes: >Um... I think I would go with Lenny's version (No offense intended to Mr. >Becker) merely because of the sort/discard option he suggests. However, >I don't recall seeing the word 'backup' in either of the follow-ups! >I am currently running the Maxtor 2190's (Thanks Thad) and am also in >agreement with Lenny on this: the seek time is not very good. All that >space on the other hand... Well, here comes the word 'backup'. WARNING: If you LOATHE the UA hit 'n' right now. I don't like it much either, but . . . How come no one has mentioned doing this: A) do a full backup via the UA, B) swap old HD with new HD C) use the diagnostics diskette to format the new drive, D) load Foundation Set E) then use the UA to do a full restore. It has its disadvantages (i.e. it's SLOW, it's REALLY S...L...O...W...) but it works and it's simple. And since it uses cpio, I assume that it will also defragment the disk as one would like. This solution also works if you don't happen to have any means of mounting the extra hard drive. Did I mention that it takes some time? It does have some advantages. You don't have to worry about the loader, or the swap space, AND you have a backup of your system when you're done. This worked for me when I bought a bigger drive. Of course, it took about 5 hours just to migrate from a 20 meg to a 40 meg drive. Most of the time was spent in the scintillating process of backup and restore. The rest in the equally stimulating process of formating. Ok, so it's not a good suggestion if you can mount the extra drive, but if you can't it will work fine. - sparkie