smd@rehder.larc.nasa.gov (Steve Dahmen) (12/04/90)
Thanks all those at SGI and elsewhere for filling me in on the problem. Due to the teeny default root partition, the /tmp filled up too quickly when compiling large files. The kernal swap caused the problem because we saved the old kernal in / ! One interesting comment I heard was that perhaps the root partition default size needs to be upped by the distribution folks at SGI. For whatever its worth. Thanks again! -- -*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-* Steve Dahmen, Systems Analyst (804) 864-4519 (W) M/S 365, NASA Langley Research Center Hampton, Virginia 23665 smd@rehder.larc.nasa.gov
wicks@DCDMJW.FNAL.GOV ("Matthew J. Wicks") (12/04/90)
>>Thanks all those at SGI and elsewhere for filling me in on the >>problem. Due to the teeny default root partition, the /tmp filled up >>too quickly when compiling large files. The kernal swap caused the >>problem because we saved the old kernal in / ! >> >>One interesting comment I heard was that perhaps the root >>partition default size needs to be upped by the distribution folks at >>SGI. For whatever its worth. Another solution (one that we use here) is to have a separate mounted file system for /tmp. We do the same for /usr/tmp Matt Matt Wicks Fermi National Accelerator Laboratory 708-840-8084 wicks@fnal.fnal.gov
mitch@sgi.com (Thomas Mitchell) (12/28/90)
In article <9012041407.AA09789@dcdmjw.fnal.gov> wicks@DCDMJW.FNAL.GOV ("Matthew J. Wicks") writes: >>>Thanks all those at SGI and elsewhere for filling me in on the >>>problem. Due to the teeny default root partition, the /tmp filled up >>>too quickly when compiling large files. >>>One interesting comment I heard was that perhaps the root >>>partition default size needs to be upped by the distribution folks at >>>SGI. For whatever its worth. > >Another solution (one that we use here) is to have a separate mounted >file system for /tmp. We do the same for /usr/tmp > Merry Christmas == Disk space for temp files is a common problem. It turns out that there are some generic Unix/Irix 'solutions'. There is a library function 'tmpnam' which is used to generate unique file names for scratch files (see man tmpnam). This library function is used by lots of programs, including cc and f77. By default it places temp files in the /tmp directory. Commonly this is too small for modern systems. 'tmpnam' understands an environment variable TMPDIR in the user's environment, whose value is the name of the desired temporary-file directory. I commonly set my TMPDIR to $HOME/tmp because my home directory lives on a disk with bunches of space. One of the strengths of Unix/Irix is the shared code used in program after program. 'tmpnam' is one of these commonly shared functions. If we all had source we would know this. For those of us without source I use the program 'strings' on the binary and egrep for TMP|TMPDIR|TEMP in the output. Finding TMPDIR tells me the program is using 'tmpnam' (99%). Another common user of /tmp is vi. 'vi' does not use 'tmpnam'. It can however be directed to place its scratch copy (set directory=/usr/tmp) in a roomy place in a number of different ways. 'vi' understands the environment variable EXINIT and also looks at the file $HOME/.exrc or ./.exrc. The rules that govern which rc file or environment variable 'vi' uses are documented in the 'ex' man page. Since vi is a link to ex this makes sense but it is not obvious. In my $HOME/.exrc I set directory=/usr/tmp I do other explicit things for my root user to avoid 'dumb traps'. Traps, which are the reason for ex/vi's complex start up rules. -- -- Thomas P. Mitchell -- mitch@sgi.com or mitch%relay.csd@sgi.com "All things in moderation; including moderation."