Jump to content

Obsolete:IPMI on Sun servers

From Wikitech
(Redirected from IPMI on Sun servers)

For IPMI on the toolservers see [1] (main page Toolserver). (The page Toolserver/Use the SP for remote control is probably out of date.) The document you are reading covers all the other Sun hosts.

Sun's IPMI is either the Integrated Lights Out Manager (ILOM) or the Embedded Lights Out Manager (ELOM), depending on the hardware platform. You can use this to power up systems and to retrieve information about the hardware, in theory. For ILOM, we are running version 2.0.x on some hosts and 3.0.x on others; for ELOM, we are running 4.0.x on the hosts that have it, i.e. the search## hosts.

Logging in

Our Sun servers are currently either in the esams location or in pmtpa. In order to access the IPMI interface on these hosts you choose one of the two methods according to the location:

Hosts in pmtpa (includes ms1 through ms6, db11-db30, searchidx1 and search1-12. (Currently the interfaces on db30 seem to be inaccessible, who knows why.)

  • Be on a host in pmtpa
  • ssh -l root xxx.mgmt.pmtpa.wmnet
  • password is the same as the one used for access to the drac interface

Hosts in esams (ms6)

  • Be on a host in esams
  • ssh -l root xxx.ipmi.esams.wikimedia.org
  • password is the same as the one used for access to the drac interface

This will drop you at the -> prompt.

Basic commands (ILOM)

  • exit -- leave the command line interface and exit the session
  • show /SP -- show information available from the serial port
  • start /SP/console -- connect to the host on the serial port (should give you the regular Solaris or Linux login prompt)
    exit the console session with ESC ( (it will tell you this when the session starts as well)
  • start /SYS -- power on the host
  • stop /SYS -- power off the host
  • restart /SYS -- power cycle the host

Troubleshooting (ILOM)

If the ILOM gets itself into a strange state, for example if you try 'show /SYS' it says 'No such target', you can try the following.

  1. reboot the host from Solaris
  2. reset /SP -- sometimes this will clear up the problem
  3. if you have an old version of ILOM, update it.
  4. all else has failed? shut down Solaris and ssh into the ipmi interface with the username sunservice. Use the drac password to log in. Issue the command 'reboot'. Now, wait several minutes and reconnect as root and check the ILOM again.

Note: the sunservice account gives you a linux command line (that's the basis of the ILOM). So you can poke around... but you can also break things. Be cautious.

Updating the ILOM

  • Set up a tftp server on a host in the same datacenter. ms4 is currently running tftpd for pmtpa.
  • Download the copy of the ILOM firmware you need for your platform from sun and put it on your tftp server.
  • Log into the ipmi interface of your host as root.
  • Give the command
load -source tftp://ip.address.to.server.here/ILOM-version-platform-etc.pkg

If you see 'tftp: timeout' several times and then 'upgrade failed', either your tftp server is set up incorrectly or you have the wrong filename, *or* you can't reach that server from your host. If you see a series of dots, it's working. You should eventually see 'Firmware update is complete. ILOM will now be restarted with the new firmware.' Wait a few minutes and reconnect. The message after login should show you the new version number.

More information

See also the Platform-specific documentation describing each of the various Sun hardware models and how LOM behaves on them.

For more information, see the ILOM 2.0 User guide or Sun Integrated Lights Out Manager (ILOM) 3.0 Documentation.