jgreco@archimedes.math.uwm.edu (Joe Greco) (10/13/90)
We've got a nice new RISC System/6000 model 530 here at UWM, which is on our local ethernet. On the same ethernet is our RT and about 15 6152 workstations, strung together with NFS. The RT has 330 Mb in it, and it's starting to get a little cramped. NFS has worked flawlessly to this point between the RT and 6152's (and even another RT on campus), so I figured that I would just export a directory on the 6000 and have a lot of megabytes of space to compile TeX in. I'm exporting /u and /usr on the 6000. The user directories are in /u. That's where I've got my temporary directory, too. I can successfully mount /u on the 6152's and read to my heart's content (at least so far), but if I try any major writes, I get locked in a NFS disk write on the 6152 (ps shows 'waiting for device I/O') and eventually I'll get "NFS server thales not responding... still trying" I've tried about six different configurations with /etc/fstab, all of which work fine from the RT. I've varied the block sizes, I've altered the retries and timeouts, I've tried soft vs hard.... I'm pulling my hair out. Nothing works. I finally set "intr" and can ^Z out of a task. When the task is suspended, I can do a "pwd" and suddenly I'll get "NFS server thales OK", but the instant I go back to the job, it starts to die again. It reads just fine but will write very little. I did *NOT* use SMIT to set up the NFS exports from thales, although they have been checked with SMIT and all appears fine. I simply created an /etc/exports file to my pleasing. Could this be a problem? Anyone else have a 6000 running NFS exports? Suggestions? Comments? ... Joe ------------------------------------------------------------------------------- Joe Greco - University of Wisconsin, Milwaukee - Department of Mathematics jgreco@archimedes.math.uwm.edu USnail: Joe Greco Voice: 414/321-6184 9905 W. Montana Ave. Data: 414/321-9287 (Happy Hacker's BBS) West Allis, WI 53227-3329 ICBM: 43 05 20 N 87 53 10 W #include <witty_and_humorous_saying.h> Disclaimer: I don't speak for the Math Department, the University, or myself.
peter@dbaccess.com (Peter A. Castro) (10/17/90)
in article <6936@uwm.edu>, jgreco@archimedes.math.uwm.edu (Joe Greco) says: + + We've got a nice new RISC System/6000 model 530 here at UWM, which is on our + local ethernet. On the same ethernet is our RT and about 15 6152 + workstations, strung together with NFS. The RT has 330 Mb in it, and it's + starting to get a little cramped. NFS has worked flawlessly to this point + between the RT and 6152's (and even another RT on campus), so I figured + that I would just export a directory on the 6000 and have a lot of megabytes + of space to compile TeX in. + + I'm exporting /u and /usr on the 6000. The user directories are in /u. + That's where I've got my temporary directory, too. I can successfully + mount /u on the 6152's and read to my heart's content (at least so far), but + if I try any major writes, I get locked in a NFS disk write on the 6152 (ps + shows 'waiting for device I/O') and eventually I'll get "NFS server thales + not responding... still trying" + + I've tried about six different configurations with /etc/fstab, all of which + work fine from the RT. I've varied the block sizes, I've altered the + retries and timeouts, I've tried soft vs hard.... I'm pulling my hair out. + Nothing works. I finally set "intr" and can ^Z out of a task. When the + task is suspended, I can do a "pwd" and suddenly I'll get "NFS server thales + OK", but the instant I go back to the job, it starts to die again. It reads + just fine but will write very little. + + I did *NOT* use SMIT to set up the NFS exports from thales, although they + have been checked with SMIT and all appears fine. I simply created an + /etc/exports file to my pleasing. Could this be a problem? Not really. There could be another problem which should have been in your 6000 documentation concerning networking with the RS/6000 and a PC/RT. This concerns the size of an Ethernet packet. On the RTs this size is about 1518 bytes in length (inetlen). On the RS/6000 the length of a packet is 1500. This can lead to problems that I'm sure you can imagine on your own. Look in the file /etc/net on your RTs, and you'll see a stanze like the following: * @(#)net 7.7 88/11/18 14:30:44 * This file contains a stanza for each adapter card (device) that * TCP can use. The stanza describes the characteristics for that * device. [comments deleted ..] net0: netaddr = 128.32.152.2 inetlen = 1518 r_inetlen = 1518 subnetmask = x0ff00 Now, just change the 1518 (both of them) to 1500 and things should be better. Hope this helps. + + Anyone else have a 6000 running NFS exports? Suggestions? Comments? Yes, I'm running a system that exports and imports NFS to/from an RS/6000 530 to an RS 6150. + > ... Joe + + ------------------------------------------------------------------------------ + Joe Greco - University of Wisconsin, Milwaukee - Department of Mathematics + jgreco@archimedes.math.uwm.edu USnail: Joe Greco + Voice: 414/321-6184 9905 W. Montana Ave. + Data: 414/321-9287 (Happy Hacker's BBS) West Allis, WI 53227-3329 + ICBM: 43 05 20 N 87 53 10 W + #include <witty_and_humorous_saying.h> + Disclaimer: I don't speak for the Math Department, the University, or myself. -- Peter A. Castro INTERNET: peter@dbaccess.com // //| c/o DB/Access Inc. UUCP: {uunet,mips}!troi!peter // //|| 2900 Gordon Avenue, Suite 101 FAX: (408) 735-0328 \\ // //-||- Santa Clara, CA 95051-0718 TEL: (408) 735-7545 \// // ||
jgreco@archimedes.math.uwm.edu (Joe Greco) (10/20/90)
In comp.sys.ibm.pc.rt article <772@troi.dbaccess.com>, peter@dbaccess.com (Peter A. Castro) wrote: :in article <6936@uwm.edu>, jgreco@archimedes.math.uwm.edu (Joe Greco) says: :+ I'm exporting /u and /usr on the 6000. The user directories are in /u. :+ That's where I've got my temporary directory, too. I can successfully :+ mount /u on the 6152's and read to my heart's content (at least so far), but :+ if I try any major writes, I get locked in a NFS disk write on the 6152 (ps :+ shows 'waiting for device I/O') and eventually I'll get "NFS server thales :+ not responding... still trying" .... : Not really. There could be another problem which should have been in your : 6000 documentation concerning networking with the RS/6000 and a PC/RT. This : concerns the size of an Ethernet packet. On the RTs this size is about 1518 : bytes in length (inetlen). On the RS/6000 the length of a packet is 1500. : This can lead to problems that I'm sure you can imagine on your own. Look : in the file /etc/net on your RTs, and you'll see a stanze like the following: Hey, all right! Except I'm running AOS, not AIX (I assume this is an AIX-fix). However, you did hit the nail on the head - I set the read and write sizzes in /etc/fstab to be 1024 or 1280 or some nice even number like that. Suddenly things work (albeit somewhat more slowly) with the RS/6000. I had experimented with raising the sizes and hadn't lowered them sufficiently to discover this. I'm using 4096 for inter-RT NFS (not sure if this is even valid, but it does seem a bit faster than 2048). ... Joe ------------------------------------------------------------------------------- Joe Greco - University of Wisconsin, Milwaukee - Department of Mathematics jgreco@archimedes.math.uwm.edu USnail: Joe Greco Voice: 414/321-6184 9905 W. Montana Ave. Data: 414/321-9287 (Happy Hacker's BBS) West Allis, WI 53227-3329 ICBM: 43 05 20 N 87 53 10 W #include <witty_and_humorous_saying.h> Disclaimer: I don't speak for the Math Department, the University, or myself.