Inventory of how the Manoa cluster is used and some details of the hardware.

Nodes criticalness

When necessary, the nodes labeled 1 can be shut down first. Then, the nodes labeled 2, and so on...

* 1: Can be shut down. Not necessary for operations:

  • poi00 (offline now ipp000, master, nkbf0, alala)

* 2: Could be shut down but necessary for full operations.

  • ipp003,

* 3: Shouldn't be shut down. Absolutely necessary for operations.

  • ipp001, ipp002, ipp022, ippops1, ippops2,

* 4: Shouldn't be shut down unless data shuffle not needed

  • ippb04, ippb05

All systems should have sensors available to check temperatures. Temperatures also available on the ganglia plots http://ganglia.pan-starrs.ifa.hawaii.edu/?c=IPP%20Development

ipp000

ipp001

  • dumps of gpc1/ippadmin/isp, daily, every 4 hours
  • mysql data base for ifa pstamp request and development test data store
  • ippMonitor/transients/ippBlog
  • proxy
  • has idl
  • mysql slave
  • 4 cpus/ 8gb
  • 8 TB

ipp002

  • has some files from rsync server (the big threepi db)
  • svn/trac
  • 4 cpus/ 8gb
  • 8 TB

ipp003

  • rsync server
  • proxy
  • 4cpus/ 8gb
  • 10 TB
  • transients DB/name server

ipp022

  • rsync server
  • has idl
  • 8 cpus/ 16gb
  • 20 TB
  • datastore (IfA users postage stamps request) -- Very small. Bill says could be moved to ipp001
  • ipp build for running data store and IFA postage stamp request submission scripts. Bill says it could be moved.

ippops1

  • ipp datastore/pstamp proxy to ippc17

ippops2

  • ganglia monitor server

master

  • IPP gentoo OS (portage and driver) development

poi00

  • IPP NIS secondary

nkbf0

  • rsnapshot server (HW raid controller flaky, ARC-1170 PCI-X)

alala