In message <m0ydBjF-000035C@druid.net>, D'Arcy J.M. Cain writes:
>I suppose I could increase map entries (how?) but I suspect that would
>just push the problem a little further back.  Anyone have any ideas?
>Would UVM solve this problem?  Is UVM baked enough for production
>systems?

Well, I can't say for sure, but the first thing I do with any resource
shortage is bump maxusers.

>Oh, and if you have had a NetBSD system running for more than a few months
>can you send me your uptime output?

Sadly, my NetBSD system went down last weekend when a power outage turned
out to be about 22 hours longer than our UPS.

Apart from that, well, honestly, I haven't seen a NetBSD box crash since...
hmm.  I think it would have been before we got to 16 MB of RAM on the A3000.
Apart from that, the only reason my boxes go down is hardware.  I'd send
uptime outputs, but I've had a lot of hardware changes recently.  :)

>I need to have something to show
>people when they shove those FreeBSD uptimes under my nose like as if
>it proves something.

It does!  It proves that somewhere out there there's a FreeBSD box that's
running.  If you want, I can give you a 'who' output from a BSD/OS 1.1 system
up over a year.  (No offense meant to BSDI, but BSD/OS 1.1 was *not* the most
stable release ever...)

For a good laugh, tell them that you're using NetBSD, not FreeBSD, because
FreeBSD is free software, and not supported.  :)

-s