Shutdown priorities

The following list puts priorities on computers, equipment and other items of interest. Computers in racks, which will stay powered up, should be unplugged!

Shutdown plan

level what / detailed ordered list
1 no standard computation, quick to recover
1.1 ra01..ra10
1.2 n0422..n0588 (Lenny)
1.3 n0589..n0840
1.4 n1345..n1662
1.5 E@H mirror
2 harder to recover
2.1 obsolete Racks, obsolete LCPs, Umluftk├╝hler
2.2 gpudev{1,2,3,4}, gpu0{1,4,5,8}, PS3
2.3 mds2, s00, FAI-lenny, monitor-nodes
2.4 E@H dev machines (PowerMac, MacPro, ...)
3 basic cluster service down
3.1 bob, scr01, einstein-dl2, QFS-clients, DDN
3.2 cfengine, kerberos, unknown(mgmt/25), epilog, cornell, p-kvm, condorslave, external01
3.3 einstein-home, d3{3,4,6,7,8,9}
3.4 file server (possible except d03 for now)
4 limited access to cluster
4.1 postfix, titan1, titan2, titan3, atlas3, atlas4
4.2 switch off condor on atlas1, then condormaster, backup server
4.3 vm, n0, LDR, geosegdb
4.4 switch off 8 of condordev's pool
5 no user access possible
5.1 thumper (not s12), MDS server, 6780, atlas2, fluffier, bute
6 going unsafe
6.1 nut
6.2 by-pass UPS
7 second to last barrier
7.1 condor dev pool, condordev, s12, orion
7.2 einstein-wug, einstein-db
7.3 einstein-abp1, einstein-dl
7.4 atlas1, fai
8 no network, no monitoring, no access, no nothin'
8.1 darkness (switching...)

Core/External network

  • Cisco switch
  • Woven/Fortinet switches
  • HP mgmt switch

Head node/condor - most important ones

  • atlas1 (web server for LVC meeting)
  • condormaster?
  • monitoring server/nodes
  • n0
  • VM
  • LDR
  • FAI

I think these are all needed
  • einstein-wug
  • einstein-abp1
  • einstein-dl
  • condordev
  • orion
  • very few condordev nodes
  • s12

These might not be needed right now (ask Bernd/Olli)
  • einstein-db
  • einstein-dl2

$HOME servers

  • s02..s11/s13
  • HSM
  • MDS/MDS2
  • QFS clients

headnodes/othe rmachines - not so important

  • atlas2/3/4 (atlas4 should be more available than atlas2)
  • atlas2 is NAT gateway, but needed?
  • titan1/2/3

misc (mgmt) nodes

  • nut
  • bob
  • gpudev
  • gpu
  • eahm
  • scr01
  • Macs/Playstation
  • file server (d01.."d40")

most dispensable

  • compute nodes
  • s00
  • gpuXXX *

MISC

  • Any Rittal racks should be shut down which are not needed
  • Air conditioning
  • UPS?

-- CarstenAulbert - 14 Mar 2012
Topic revision: r4 - 15 Mar 2012, CarstenAulbert
This site is powered by FoswikiCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding Foswiki? Send feedback