parker@mpre.UUCP (Ross Parker) (10/07/87)
I recently attempted installing Ultrix 2.0 on one of our Microvax-II's. The pertinent bits of the system configuration are as follows: - Microvax-II with 16 Mb memory - Fuji Eagle on Emulex controller as boot device (ra0) - RA-81 on KDA-50 (ra4) - tk-50 cartridge tape I installed Ultrix 2.0 on this machine, and very quickly discovered that it ran like a slug... *much, much* slower than with Ultrix 1.2. After a little looking around, I found that it was paging correctly from the boot device (the eagle), but was not paging on the ra-81. The kernel had been correctly configured for paging from both drives, both b partitions were set up in /etc/fstab as swap partitions, and a 'swapon -a' is in /etc/rc. I attempted to do a 'swapon /dev/ra4b' (the ra-81's b partition), and got '/dev/ra4b: Mount device busy', which indicates that the swapon -a had been successful. Additionally, I could access the RA-81 in every other way. I could mount, read, and write all partitions that I had set up. There is one thing that I can think of that is suspect. During the installation, I found out that Ultrix 2.0 is a lot larger than Ultrix 1.2, and effectively filled up the root partition with no room for even a second copy of vmunix. I had therefore increased the size of the 'a' partitions on both drives to twice the 'standard' size, and while I was at it, I also doubled the size of both 'b' partitions. I called DEC support about the problem, and their reply was basically that 'that was the only thing that they could think of that was different from the normal setup, so that must be the problem'. I don't consider that a very good answer. I'd rather have someone say "I *know* that's what the problem, because ... ... ..." Can anyone help? I sure hope so! -- Ross Parker ...ihnp4!alberta!ubc-vision!mprg!parker | Microtel Pacific Research Ltd. | Every silver lining Burnaby, B.C., | has a touch of grey... Canada, eh? |
finegan@uccba.UUCP (Mike Finegan) (10/23/87)
In article <1315@mpre.UUCP>, parker@mpre.UUCP (Ross Parker) writes: > > I recently attempted installing Ultrix 2.0 on one of our Microvax-II's. > The pertinent bits of the system configuration are as follows: > > - Microvax-II with 16 Mb memory > - Fuji Eagle on Emulex controller as boot device (ra0) > - RA-81 on KDA-50 (ra4) > - tk-50 cartridge tape > > I installed Ultrix 2.0 on this machine, and very quickly discovered that > it ran like a slug... *much, much* slower than with Ultrix 1.2. ... > The kernel had been correctly configured for paging from both drives, both > b partitions were set up in /etc/fstab as swap partitions, and a 'swapon -a' > is in /etc/rc. ... > I found out that Ultrix 2.0 is a lot larger than Ultrix 1.2, and effectively > filled up the root partition with no room for even a second copy of vmunix. > I had therefore increased the size of the 'a' partitions on both drives to > twice the 'standard' size, and while I was at it, I also doubled the size > of both 'b' partitions. I called DEC support about the problem, and their > reply was basically that 'that was the only thing that they could think of > that was different from the normal setup, so that must be the problem'. I have recently been involved in installing Ultrix 2.0, and felt that the `advanced' installation was less than advanced. If you install all the supported software it just about fills the / file system, and if you install Decnet, and all the optional software, it is over 100% (with 10% margin) when you first go multi-user! The answer is the mystical precedence of an 8 Megabyte root partition. Okay, but with +100% there are tons of soft links to /usr/etc - at least move enough to start below 100%. Anyhow, we have 1 RA81 with both modified rra0a and rra0b partitions (as well as the rest - something I had to REDO because of the lack of freedom in the installation script - forced default partitioning), with 12M of root and 16M of swap (2 times our 8M "core"). I would think that this was not your problem. Our system is the same (on a 750 its hard to tell if it's faster), definitely no worse. In the config files, you can not only say swapon ... but also swapon ... size `sectors' . Did you change this to take advantage of your new swap partition size ? I don't remember if you give the starting sector as well. I think there is an example in the manual, or in the Berkeley supplemental manuals. Since we only have one disk, I don't know how that line would change for two disks. Hope this helps, Mike Finegan ...!{hal,pyramid,decuac,mit-eddie}!uccba!finegan
chris@mimsy.UUCP (Chris Torek) (10/26/87)
In article <1263@uccba.UUCP> finegan@uccba.UUCP (Mike Finegan) writes: >... Anyhow, we have 1 RA81 with both modified rra0a and rra0b partitions >(as well as the rest - something I had to REDO because of the lack of >freedom in the installation script - forced default partitioning), with >12M of root and 16M of swap (2 times our 8M "core"). [We use ~30 megabyte roots, which is perhaps a wee bit excessive, but quite effective!] >... In the config files, you can not only say swapon ... but also >swapon ... size `sectors' . Did you change this to take advantage of >your new swap partition size? This should not be necessary. Starting with 4.2BSD, block device drivers have a `d_psize' partition sizing routine, so that the swap configuration code can find the sizes of the swap partitions. Putting a size into the configuration file overrides the actual partition size, but the only workable thing you can do with this is to decrease the size of the swap region. -- In-Real-Life: Chris Torek, Univ of MD Comp Sci Dept (+1 301 454 7690) Domain: chris@mimsy.umd.edu Path: uunet!mimsy!chris