richv@hpinddu.cup.hp.com (Rich Van Gaasbeck) (06/21/91)
I'm finding more and more evidence that CNEWS expects everthing below /usr/spool/news to be on the same file system. My first clue is that the space determining program (I forget the name, spacefor?) as written only looks at the space on the filesystem passed as a argument (and none below it). A fellow news user says there is a second problem, CNEWS hardlinks crossposted articles, which won't work across file systems. Am I right? Are the choices 1) use ONE very large disk, or 2) expire articles fast enough to fit whatever ONE disk (really filesystem) you have and not 3) use several disks (filesystems) to hold /usr/spool/news. Richv richv@hpinddu.cup.hp.com
henry@zoo.toronto.edu (Henry Spencer) (06/25/91)
In article <RICHV.91Jun21171136@hpinddr.cup.hp.com> richv@hpinddu.cup.hp.com (Rich Van Gaasbeck) writes: >I'm finding more and more evidence that CNEWS expects everthing below >/usr/spool/news to be on the same file system. My first clue is that >the space determining program (I forget the name, spacefor?) as written >only looks at the space on the filesystem passed as a argument (and >none below it). A fellow news user says there is a second problem, >CNEWS hardlinks crossposted articles, which won't work across file >systems. Your fellow user hasn't looked quite closely enough, since C News will try to make a symbolic link if the hard link fails. (Expire does not support this perfectly unless you give it the -l option, mind you.) That aside, you've got the right idea but not expressed quite precisely enough. C News makes one space check before processing a batch. That check doesn't necessarily have to check only one filesystem, although that's the way it works as supplied. There is no reason why a custom spacefor couldn't check multiple filesystems and take the minimum, or something like that. The reason for doing one check per batch is simple: space checks are potentially quite expensive. Note also that, as documented, the space checks are not intended to let you run right up to the ragged edge, only to head off gross problems. That being the case, one check per batch seems a reasonable approach. -- "We're thinking about upgrading from | Henry Spencer @ U of Toronto Zoology SunOS 4.1.1 to SunOS 3.5." | henry@zoo.toronto.edu utzoo!henry