1) Is it true that it is safe to delete page files without
going through the deinstall command under OpenVMS AXP 1.5
and OpenVMS 5.5-2? If so, since when (from which version of
VMS) did Dec made it so?
It's always been safe to delete *a pagefile that was not in current use* -
i.e., made available to VMS SYSGEN's INSTALL/PAGEFILE command. Most page-
files are installed during system startup, PAGEFILE.SYS automatically.
What's dangerous is deleting a pagefile while low-level VMS code has access
to it as a pagefile. Since that access doesn't go through the file system,
in the past the file system had no idea that the file was in use and would
happily delete it. Then the file system would use the now-free space for
files, while low-level VMS code continued to page to it. Boom.
Starting with, I think, V5.0 of VMS, VMS "opens" pagefiles through the file
system. It still doesn't use the file system to access pagefiles, but because
the file system now "knows" that the file is in use, it won't let you delete
it. This is just a safety measure, keeping you from doing something stupid.
Note that if you manage to over-ride the file system and actually delete a
pagefile while it is in use, you are still safe: The file system will mark
it deleted, but won't actually reuse the space because someone (VMS) will
still have a channel open on it. Since that channel probably won't be closed
until the system crashes (whether in a actual crash or at the end of a
shutdown), the file space will remain allocated to a file marked for delete.
You can recover it by doing an ANALYZE/DISK/REPAIR.
Also new in, again I think it was V5.0, was the SYSGEN DEINSTALL/PAGEFILE
command. This simply tells VMS to stop using the pagefile and "close" it,
at which point it can be deleted safely.
2) I have a couple of vaxes and an axp using the HSC cluster/drives.
Everytime, I reboot the axp, it always does a rebuild on each of
the HSC drives (for every mount). Is it safe to avoid it by adding
the /NOREBUILD qualifier to the mount command in the axp startup
file? What the consequences?
It costs you some space on the disk. The VMS filesystem algorithms are
designed to "fail safe". For example, in a cluster each processor keeps a
cache of blocks that it knows are free. It will allocate those to files it
creates. To keep other processors from trying to use those blocks, it marks
them as allocated in the on-disk bitmap, even though they really aren't yet.
During a normal dismount of the disk, the processor can "return" any pre-
allocated blocks by simply marking them free again on the disk. Should the
system not properly dismount the disk, the blocks remain "allocated", but to
no file. The disk rebuilding process will notice this and fix it.
You can do a rebuild at any time with SET VOLUME/REBUILD. One common practice
is to mount volumes /NOREBUILD, then schedule a batch job to rebuild the
volume at a "quiet" time (say, 3:00 AM). Alternatively, SET VOLUME/REBUILD
is more or less a no-op if the volume doesn't need rebuilding. (I think it
will still force the processor caches to be flushed, but that's a pretty
minor effect.) So no great harm is done if you rebuild your disks every
morning at 3:00 AM.
-- Jerry