• ComputingResources
• TroubleShooting

• Out Of Date

  This page contains material that may no longer be current.

(Probably a good idea to keep a printed copy of this somewhere, just in case the wiki can't be accessed)

UPS-activated shutdown

diagnosis: message on smithers shortly before shutdown, claiming UPS is low on power. Shortly after, computer activated shutdown procedure.

• 2006-12-04 -- smithers was shutdown, due to UPS claiming that it was low on power

• 2006-12-05 -- resolution (DavidHall):

  • reboot (turn on) smithers

[Note: after a kernel upgrade, and changing BIOS settings to start up with power on, turning on smithers after a power shutdown is less likely to be necessary]

Interface DNS server down

diagnosis: cannot get DNS resolution for interface.org.nz. For example, ping user.interface.org.nz doesn't work, but ping 130.195.11.127 does.

• 2006-12-19 -- no resolution for interface.org.nz domain

• 2006-12-19 -- resolution (DonaldGordon):

  • attempt to ssh into Interface DNS server
  • discover that this is not possible (hint: if you can't ssh to the machine, that would tell you why it's not acting as a DNS server.)
  • ssh into the Xen machine that hosts unibus
  • "xm console unibus" as root
  • discover that it's kernel panicing
  • "xm reboot unibus"
  • discover this doesn't work
  • "xm create -f /etc/xen/auto/unibus" instead (creates a xen domain from the named configuration file, i.e. boots up unibus after you discovered that reboot made it shut down but not restart)
  • discover that the machine now boots and that DNS works

Mail and SWANS transparent proxy broken

diagnosis (< 2007-1-1): mail to Interface mailing lists bounces, access to port 80 via SWANS to the outside world fails

• 2007-1-12 -- Fixed by DonaldGordon: '/var on unibus was full (which also broke the proxy) but now it's not.'

• 2007-1-22 -- Fixed by LoRne: restarted mailman process on unibus, assumed to be dead a result of /var filling up.

vtund / excess processing swamping/freezing interface servers

diagnosis: very difficult (if not impossible) to ssh into smithers, CPU load average on burns (via 'uptime') exceeds 5

• temporary fix that seems to work: disable vtund
  • xyzzy has stated that this is bad, and will break access to the SWANS router. This breakage has not yet been observed
    • if you kill vtund, then then the AP suddenly won't be able to connect to the virtual SWANS ethernet
    • there may be some redundancy in this -- smithers seems to be running a vtun server, which may stop SWANS breaking

• solving this might require a better understanding of HowSwansWorks