Bladewatch.com

Upgrading legato need not be difficult and should not require a reboot.

I’ve been upgrading the legato client on windows servers, from 6.02 to 7.34, and one of the things I noticed was that 7.34 would insist you uninstalled the old version, however, this worked:

• Upgrade from 6.02 to 6.13
• Now upgrade from 6.13 to 7.34
• Most worked fine without a reboot. You’d be best to try this on something that cam be rebooted in case it asks, I think a few did want rebooted.

Best practice would suggest you uninstall and cleanly install version 7.34, also note that when you do the upgrade you are installing the correct options (client only etc).

Share and Enjoy

Who should pay to upgrade fibre cards? IT or the application teams?

I was having a chat with one of the project managers for a relatively small hedge fund in the London area. He was involved in a SAN storage upgrade project, basically the IT team are moving their storage from one storage rig or device to the other. 70% of the servers are fine, running relatively new Emulex fibre cards and these can be migrated to the new san storage unit.

• However, there is a problem which is threatening to derail his project:
• Who pays to replace the fibre cards that are not supported? He has budget but not to buy new cards.
• If he buys them does that set a precedent for IT to be retrospectively paying to upgrade components as they reach end of life.
• What is the deemed life span of a fibre card and is that a component that is a business cost or a business as usual cost?
• What are the defined terms of support and are they agreed?

We’ll have to see, I believe conversations are being had, sign off approved, these concepts can seem unimportant and in a million pound project a few thousand pounds could be seen so, however it’s often the little things that cause the most issue. A few sample statements I got told about:

1. “How long have you know my cards are out of support”
2. “What actions have you taken to report that my cards are out of support”
3. “What kind of testing and due diligence will need to be actioned before ordering new cards”
4. “What deemed outage will there be to upgrade the cards?”
5. “If we’re already out of support, is it actually an issue to move to another out of support configuration?”

Share and Enjoy

Writing about my obsession with firmware and drivers.  The reason I routinely write about firmware and drivers is that it’s good best practice.  Your vendor/service provider will want you to be running the latest supported version of the drivers and firmware to ‘be in support’. With that in mind it’s good general practice to be keeping both up to date.

Keep in mind the following:

• Upgrading the drivers might necessitate a driver upgrade – read the instructions or if in doubt ask
• Upgrading the firmware might need to be done offline (rebooting the server first then booting from a device with the image)
• Drivers/layered software might have dependent components – the network driver and card software are examples or the management devices/system peripherals.
• Firmware could affect the operating system, on Windows NT (apologies for repeating myself), one of the roll up packs required your array controller firmware and drivers to be a specific version
• Drivers need to be tested – new drivers need to be tested on your environment for functionality as does the firmware.
• In the fibre card world, your firmware, your drivers and the fibre switches/uplinks might need specific options set and drivers/components loaded to work.
• Firmware upgrades might need to be done in incremental steps, for example 1.4 to 1.92 on the HP ILO firmware should work but is quite a change in one go.
• It’s the lowest level, your firmware underpins your operating system, get your operating system and your platform, optimized, secured and operating as it should and everything else will follow.

Share and Enjoy

I got asked by a friend today to write five ways to instantly reduce your oncall/support costs:

1. Clear down unnecessary disk space – learn whatkey layered applications/operating system components are eating your disk space, so you can script or understand the quickest way to resolve.
2. Check your monitoring configuration – it’s so easy for an alert to be generated due to poor configuration – let’s not forget a call may only last five minutes but cause a disruption/cost of hours.
3. Configuration of the server – even simple things like page file setting, location, firmware versions and driver packs can make a big difference between stability, performance and functionality.
4. Security patching/updates – installing the relevant updates and rebooting the server can make the difference between the application working and not. Operating systems are getting more resilient and reliable, but we should avoid getting in the situation where a server has been up for so long, no one knows if it will reboot ok.
5. Process – are we trapping what’s causing our calls, what aspects could be automated or scripted, what basic checks firstline can do before escalation? Do I really need to call a DataSynapse guy at 3am if one engine goes offline?

Share and Enjoy

A brief unracking guide, please note this was written on the train on my Blackberry, so you might want to adjust it.

• Confirm server is powered off.
• Disconnect cables
• Check the server is not connected to the rack or rails by any vendor specific cable management arm or device
• Remove any cable management arm/device
• Remove cables from the server and trace back to the cabinet – should we be removing the KVM and power cables? Do this carefully to avoid outage
• Remove the server by sliding it from the rails – noting that there might be metal clips holding the server in the rails.
• Remove the rails, there will be several ways to do this, check for bolts holding the rails in at either end, and for clips half way through the rails to adjust.

Share and Enjoy

I need to find another job, my contract was not renewed. This is unfortunate, as we speak I finish tomorrow, so I shall start looking, in the meantime, my blog continues. If you have any comments/suggestions on content, do get on touch, I’m always happy to hear your views.

Share and Enjoy

http://news.prnewswire.com/DisplayReleaseContent.aspx?ACCT=104&STORY=/www/story/03-31-2009/0004997547&EDATE=

AVON, Mass. and MILPITAS, Calif., March 31 /PRNewswire/ — Terascala, Inc., (www.terascala.com) a provider of high throughput, high capacity storage appliances and Appro (http://www.appro.com), a leading provider of supercomputing solutions, today announced their partnership to maximize high performance computing solutions through the combination of Appro’s Xtreme-X(TM) Supercomputers and Appro’s HyperGreen(TM) Blade Cluster solution using Terascala’s high throughput storage appliances.

Appro’s Xtreme-X Supercomputer Series paired with Terascala’s MTS 1000 Storage will offer high performance and high availability with high throughput in an easy-to-deploy redundant storage appliance providing greater performance and enhanced value to customers with high performance and high availability requirements. Also, Appro’s new HyperGreen Blade Cluster solution combined with RTS 1000 Storage solution from Terascala offers high throughput, high capacity, in an easy-to-deploy Lustre-based appliance, simplifying the management of HPC storage while improving the efficiency and I/O performance for flexible and open cluster computing deployments. Appro and Terascala together offer customers flexibility based on a common goal of helping them reduce complexity in deploying and managing their High Performance Computing solutions, while lowering their total cost of ownership.

Very cool, I’ll need to check it out, storage appliances can be a great platform for hpc solutions, particularly in those grid/caching applications.

Share and Enjoy

Vnunet

Bull Information Systems has introduced a bespoke service to help companies build a virtual machine infrastructure in their datacentres. The firm claimed that it can achieve up to a tenfold reduction in the cost and time needed to implement a solution.

Available immediately, myVMBox is a custom solution for each client according to its requirements, using the company’s NovaScale servers, Storeway storage and VMware’s virtualisation platform. Bull became a worldwide Enterprise VMware Authorised Consultant in 2008.

Bull said that its virtualisation centre of expertise evaluates the customer requirements, applying system scoping and best practice when building each myVMBox configuration.

The hardware required is then completely pre-integrated before leaving the factory, cutting the time and cost required to deploy this kind of infrastructure by a factor of 10, according to Bull.

Great news from Bull, I’ll need to read up more, anything the vendors/service providers can do to make virtualization as a platform more accessible, and successful has to be a good thing. At the same time, as end users and vendors, we need to further collaboration on best practice, evolve the processes with the technology and innovate what we can achieve with these new platforms, grid, virtualization and cloud.

Share and Enjoy

Eweek

Despite reports that IBM has taken its $7 billion offer off the table, Sun’s best option is to broker a deal to be bought by IBM, analysts say. There are not a lot of possible suitors out there for a server vendor that is struggling in an already struggling market, and remaining an independent company isn’t a good option for Sun, they said. IBM offers Sun the best chance at a good selling price, and is the best option among HP and Oracle. Though Cisco has been mentioned as a possible buyer, analysts say such a deal wouldn’t make sense for Cisco.

In the end, the move that makes the most sense for Sun Microsystems would be to be acquired by IBM.

That’s according to analysts who have been watching the rollercoaster negotiations in IBM’s attempts to buy Sun for somewhere in the neighborhood of $7 billion. IBM reportedly pulled its offer from the table April 4 after concerns from Sun executives over price.

It essentially comes down to the fact that Sun is losing a lot of money—about $209 million in the fourth quarter 2008, with more expected in the first quarter 2009—and is losing ground in a server market that itself is contracting, thanks to the global recession.

Another article about the Sun/IBM transaction, it will be interesting to see how things develop, and it’s great to get another opinion on it, we’ll have to see.

Share and Enjoy

Fujitsu

Tokyo, April 6, 2009 —  Fujitsu Limited today announced the release of the two-way, rack-mount PRIMERGY RX300 S5 and RX200 S5 servers in Japan, both using the new Intel® Xeon® processor 5500 series.

The servers incorporate enhanced power-switching and cooling technology to improve energy efficiency and help customers reduce their environmental footprint.

With the integration of Fujitsu Technology Solutions into the Fujitsu Group from April 1, Fujitsu continues to strengthen its global x86 server development and manufacturing and significantly improve the product line’s price performance. Customers will benefit from globally unified x86 server products based on the same quality standards, enabling them to build highly optimized IT environments at more affordable prices.

Fujitsu’s new two-way, rack-mount servers, the PRIMERGY RX300 S5 and RX200 S5, are powered by the Intel® Xeon® processor 5500 series. Intel released the new processor in Japan today.

Further innovation from Fujitsu of their two-way rack servers using the new Intel Xeon processors. The new Intel processors should bring further enhancements to their platforms in terms of performance and energy efficiency, I’m off to read up more.

Share and Enjoy