Logging into bogged down server

[Rick Stevens]

On Wed, 2006-04-05 at 21:08 -0400, mylar wrote:
> On Mon, 2006-04-03 at 13:40, Rick Stevens wrote:
> > On Fri, 2006-03-31 at 20:07 -0800, Harold Hallikainen wrote:
> > > For some reason, this email client deletes all the received message when I reply
> > > and then select text instead of html... When I get back home and restart the FC4
> > > machine, I'll get back on SquirrelMail. Anyways... I'm STILL trying to ssh into
> > > the bogged down server. No luck yet. I did change the DNS server to send stuff
> > > away from this server in the hopes it would eventuall finish off all those
> > > processes and start accepting my ssh login (to the IP address instead of the
> > > domain). No luck so far. In a couple days I can go hit the reset button.
> > > 
> > > Besides the modifications to configs suggested so far, I'm also thinking of
> > > adding a hardware watchdog timer to the machine so it'll reset itself should it
> > > ever do this again. I've seen a couple circuits on the web, based on a 555 timer.
> > > One of them I found looked like it could be locked up if the machine crashed
> > > during a watchdog reset. Back when I was designing MC6802 based systems, I did a
> > > watchdog timer like that and had to drive 100 miles in Kansas to hit the reset
> > > switch on a system. Watchdog timers I design now reset on an edge instead of a
> > > level so if the reset line gets stuck at either level, the watchdog will time out
> > > and reset the system. So, that's my next project! 
> > 
> > Useful.  We use remote-controlled power strips.  You can log into the
> > strip (modem or TCP/IP) and power-cycle any of the outlets on it. 
> > HIGHLY recommended for "dark data centers".
> > 
> On trick I used in a machine I couldn't log into at home that was
> connected to a UPS was to kill the power to the outlet supplying the
> UPS. This tricked the UPS into thinking there was an outage and it
> signaled the machine to go into shutdown mode. After about 20-30 minutes
> the OS and machine automatically shutdown. Then all I did was power up
> the outlet again, the machine rebooted and all was fine.

That's fine, provided the machine is "healthy" enough to see the signal
from the UPS.  That assumes the kernel hasn't gone blooey and that it's
handling interrupts properly (as most UPS thingies toggle the DTR or
DSR line to signal a power glitch).

----------------------------------------------------------------------
- Rick Stevens, Senior Systems Engineer     rstevens at vitalstream.com -
- VitalStream, Inc.                       http://www.vitalstream.com -
-                                                                    -
-          "Very funny, Scotty. Now beam down my clothes."           -
----------------------------------------------------------------------