rich@eddie.MIT.EDU (Richard Caloggero) (05/09/89)
We have a ring of 8 Apollos. Four are disked, four diskless.
One is running SR10 (bsd), the rest 9.7 (bsd). Our drives are
80meg internal winchesters.
Our primary concern right now is bringing up SR10 on all
nodes. While playing around with the one SR10 node we've got
running, it seems the new system is much less space concious than its
predecessor. For example, booting a diskless node from are SR10
node chews up about 6megs. On our 9.7 nodes, less than 1 meg is
eaten. Similar behavior is observed for logins, and displaying pads.
Any suggestions on how we might optomize our disk usage?
Another anomally we've been observing for some time now is *lost*
disk space. Every so often when we reboot one of our disked
nodes, the amount of available space on the volume seems to increase
dramatically. Where has it been hiding? Find_orphans doesn't find
it. Seems to me like a virtual memory related problem. To test
this, I wrote a little program that malloc'd some space, and
initialiued it. After allocating about 3 megs, a "df" did indeed show
a decrease of about 3meg on the volume. When I exited the program
(normal exit), "df" shows only about 2meg increase. The other meg
just vanishes. I've repeated this many times with similar results.
What's going on?
--
-- Rich (rich@eddie.mit.edu).
The circle is open, but unbroken.
Merry meet, merry part,
and merry meet again.