Farewell, old server, we knew ye well

A few weeks ago, I wrote about the curious case of the eternal server, a VM that lives on forever with no natural decline or death state. The comparison I made in that piece was to the natural lifecycle of the physical server, for — as we all know — what goes up must come down eventually.

But not all physical servers are cut from the same cloth. Every once in a while, you encounter a physical server that’s lived way beyond its normal expected lifetime and still delivers critical services day in and day out. Today, I’d like to tell you about such a server.

[ Also on InfoWorld: Paul Venezia contemplates the long life and slow death of the virtual server. | Dell, HP, IBM blade servers battle for the virtual data center ]

Way back in 2005, I was quite impressed with the HP DL585. This box was an indicator of things to come — it housed 64-bit x86 compatible Opteron CPUs that were well ahead of anything Intel was doing at the time since Intel was busy sinking billions of dollars into Itanium. It could handle four of those CPUs and up to 64GB of PC2700 RAM running at 266MHz. Add in the four 3.5-inch Ultra320 hot-swap disks up front and twin gigabit NICs out the back, and you had yourself one of the most powerful x86_64 servers on the planet at the time.

The reason I’m describing this server is that a few days ago I retired one after seven years of constant service. In an era where we tend to rip and replace hardware at the drop of a hat, this server represents a vastly different time in IT’s history. It was born in the days when Windows Server 2003 was an infant and RHEL 3 was the new game in town.

Ironically, this server wasn’t replaced because it was failing; it was replaced because it wasn’t needed. For the past five years, it had been running one form of VMware hypervisor or the other, supporting a half-dozen production VMs. As the remainder of the data center was finally transitioned from physical to virtual servers, the VMs on this box were moved to the new farm, leaving the old guy with nothing else to do. Running an ancient version of VMware ESXi with no cluster or failover support of any kind, this box held it together with aplomb. In fact, the only hardware failure encountered over its lifetime was a CMOS battery. It never even lost a disk — ever.

It was a bittersweet day in the data center. Nearby, a rack full of new Sandy Bridge-based 1U servers spun busily, each boasting 128GB of RAM and 10G interfaces, handling the myriad tasks assigned to them by a fresh VMware vSphere implementation. These replacements were vastly superior in every way to the dinosaur from another era. The commands were given and the old box’s last meaningful task was to transition its VMs to the new farm. Then, it sat idle, with nothing to do for the first time in its existence.

Rather than roughly and abruptly drop power to the old box, it was gracefully shut down and left in the rack for now. As the drives spun down and the status LEDs flickered for the last time, it was noted that the power consumption on the PDU in the rear of the rack dropped by 10 amps. Yes, the DL585 G1 was a hungry beast.

Its ILO is still connected to the network, in case it needed to be powered up for a spot check — a highly unlikely scenario. You can still log into that ILO, in fact, and see that the server is poised for action, ready for anything; it just needs a purpose again. Sadly, that’s never going to happen. Its only purpose now is to hold down the bottom of the rack and wait until that space is needed for something else, something newer. Then, it’ll find itself on the scrap heap, destined for recycling, broken up into thousands of pieces.

One can only hope that whatever was responsible for the longevity and reliability of this particular server is passed along as it becomes part of other computers somewhere down the road. If it were possible to breed computers for certain traits, this one would be a very sought-after stud — a capable performer and a proven winner.

Farewell, njesx1, may you rest in peace.