Bladewatch.com

I was speaking with Chris again today, he was after some help with his HP blades he’s playing with, he was in a data center (a bit bored) deploying Windows 2003 on some Bl460c’s. Anyway, after asking me some questions about the C class enclosure, (firmware updates and resetting the ilo on the blade using the switch on the system board to disable login), he asked me two questions which though unimportant raise two issues indirectly, so I thought I’d post them.

What’s the first server that taught you a lesson?

It was an application server, a Compaq Proliant 2500R, I can still remember the host name and where it lived in the data center, we were applying Service Pack 6a and the Security Roll Up Package which requires that you have your Compaq drivers be up to a certain level, before it was installed.

So I did the following:

• Reboot server
• Install SmartStart 5.5 (that was the latest at the time) – it went through fine and rebooted the server
• The server booted then blue screened with some kind of stop error.
• After much playing about, we discovered that upgrading the array controller firmware fixed it
  • The firmware was ancient and not supported by the driver.
• Installed SP6 + the Security Roll Up Package and that worked fine.

Ever since then it is this server, this experience that has underpinned my firmware and driver upgrade love.

The second would have been any experience with a Windows cluster, desktop heap or Emulex drivers and SP1.

What barriers to entry can you see with upgrading servers to 1GB Ethernet in the typical enterprise?

• You would need to check applications for impact of the extra bandwidth – it shouldn’t be an issue but it will need to be done for due diligence, for example any rendezvous or tibco feeds.
• Not all the servers will be 1GB ready, anything older than a DL380 G3 might not have an integrated network card that is capable of this speed:
  • What do we do with these servers – virtualize them/hardware refresh/network card to be purchased – this can be significant work just outside of changing IP/network card speed.
  • Any legacy systems, really olden days ones might only operate at 10/half or 10/full and can’t be upgraded
    • What do we do with them in terms of the upgrade project?
    • Mark for decommission?
    • Rule them out of scope?
• Upgrading might require an IP address change, as you might need to physically relocate your server on a different network or network fabric – the feeds in and out of the application will need updated as will any ip specific code references.
• The core switches might need upgraded in order to cope with this and should be checked for capacity/impact.

No related posts.

Related posts brought to you by Yet Another Related Posts Plugin.