smiddy@spam.istc.sri.com (Richard Smith) (02/02/89)
Hi Ruth, I suspect that the problem you are running into is not easily solvable. There is a per process swap limit, as well as a system wide swap limit. The per process limit, however, is hardwired into more kernel code than the system wide limit. In particular, it determines the maximum size of the page table for a process, since the in core memory map must be able to handle a process of this size if it is not swapped. If I recall correctly, this particular maximum is hard to change, especially if you do not have source. Sigh... Thanks, Richard Smith Smiddy@spam.istc.sri.com
gerber@uunet.uu.net (Andrew Gerber) (02/03/89)
>BTW, if you're intending to set up a second swap area on a new disk, >*don't* make it the first partition. Unlike regular filesystems, swap >doesn't care about the disk label in 0/0/0, and it will trash it.... I ran into the exact same problem when attempting to put dual swap partitions on our 4/110. After the "swapon" command, the entire disk (sd2) was unreadable. I had configured the swap partition as the first partition, and it had wiped the lablel. Luckily I had a printout of a dkinfo for that disk, rebuild the label with diag, and everything was OK. It's a good idea to keep a copy of the dkinfo's for all your disks around on paper, just in case. The solution is to set up the first partition containing only one cylinder, and then put your swap on sd2b. Andy Gerber
moran@warbucks.ai.sri.com (Doug Moran) (02/08/89)
Problem: 90MB swap allocated, max process allocable = 61424k This is a very familiar problem to me. If you have the Configuration Guide: Sun-3 Product Family (October 1986) look at table 3-9 on page 42. For those who don't have this guide, I will reproduce the table here: Swap Partition Max Virtual Size (MB) Address Space -------------- ------------- 6 4.3 8 8.0 (typo??) 24 16.3 40 31.7 100 61.4 200 118.7 I have queried Sun several times about ways to expand the porportion of swap space used by single process. The current response is "not a problem in 4.0". -- Doug Moran AI Center, SRI International moran@ai.sri.com