At 01:41 PM 7/24/01 -0400, you wrote:
>On 2001.07.24, Jerry Asher <[EMAIL PROTECTED]> wrote:
> > Dossy can you be more specific. Which process do you think nsvhr is
> > killing, the master server or the proxies?
>
>It's killing the process that nsvhr is loaded in -- the master server.
>
> > I am using nsvhr, but that's only in one server, and the others are using
> > nsunix.
>
>nsunix. *shudder* That's probably 50% of your woes right there, all
>in one lump.
Yeah, but not because it dies. It's because I fixed it that I have those
woes. I haven't figured out how to pawn it off on someone else yet, and
though I keep telling everyone it's only marginally faster than nssock,
people still keep filling my email with questions about it.... Um, would
you like it?
> Are the nsunix servers dying because of nsunix
>If I were to place a bet, this is where I'd put my money. Perhaps
>foolishly, but ...
>
> > I am inclined to believe the latter, as a) nsunix hasn't changed for
> > months,
>
>When did nsunix start working? Wasn't it known to be broken somewhere
>between 3.0 and now? Sorry, I haven't kept up on nsunix to know if it
>has been fixed, and my poor memory can't recall any mention of any
>fixes in the ChangeLog ...
You're teasing me? Those damned nsvhr/nsunix patches and the reactions
they have generated have been floating around for about fifteen months now,
causing grief to everyone: including me and Kriston.
> > and b) many (most?) of these crashes have occurred in the dead of night
> > when there has been no traffic to those particular servers, but at times
> > when there have been various scheduled procs running. (I can tell there
> > has been no traffic since neither the crashing server nor the nsvhr master
> > reported any traffic.)
>
>NO traffic at all, or just very, very light traffic?
No traffic. I just checked my firewall/router's messages (it logs to a
syslog daemon on the network) and there has been no traffic at all for many
of those crashes. These are public sites which is true too, that just
don't see that much traffic, unless you are googling for things like
"ACS+problemset+one+solutions" which leads to my solutions site to that
conundrum. That used to generate dozens of hits per day, now it's more
like dozens per month.
>I think there's some web search engine's spider out there that feeds
>nsvhr a HTTP request that it doesn't like, and it makes my nsd process
>fall down.
>
>If your process is dying with ABSOLUTELY no traffic, then I'd go
>looking elsewhere. It only takes a _single_ HTTP request to cause
>my nsd to die, so even at the dead of night, my monitoring tools
>pick up that my nsd died and goes and restarts it ...
>
>-- Dossy
On the otherhand, my firewall spent two weeks dieing, gack, croak, death,
from the ill-formed http headers (that contained LFs when the spec says
CRLFs) that AOLserver 3.0 was sending out for a time (until http.tcl got
fixed). aD'ers running old versions of AOLserver 3.0 would httpget my
machine with their quotations importer (from the ad problemsets) and splat,
the firewall would fall down and couldn't get back up. I couldn't believe
that a protocol as innocuous as HTML could kill a firewall, but I figure it
was an exhausted search for CRLF that failed leaving some code to poke
something out.
Jerry
=====================================================
Jerry Asher [EMAIL PROTECTED]
1678 Shattuck Avenue Suite 161 Tel: (510) 549-2980
Berkeley, CA 94709 Fax: (877) 311-8688
