I often wonder why it is so difficult for server manufacturers to make decent remote access cards. When your servers are located offsite, or require twenty-hour support, then it’s imperative that you have some way to access the machine remotely, when it requires maintenance. There’s nothing more annoying than having to travel half-way across town just to reboot a machine that you can’t get to (and I’ve done this many a time, when I was living in Amsterdam).
Modern server manufacturers seem to have a lot of trouble getting these things right. It really shouldn’t be that difficult. I started out as a sysadmin on Sun equipment, back in the mid-90s, and even back then Sun had it mostly right: a Sparc serial port is also the machine’s console. Even if the operating system is crashed, you can send a break down the line and it will fall back to the boot prom. Sure, they had their bugs – if you run a probe-scsi without resetting the prom first, it will lock up hard and need a power-cycle, but this could be worked around with a remote power switch – at least it was easy to get to, and furthermore, you didn’t need anything more than a phone line, modem and a dumb terminal to access it.
Now look at what you get from modern server manufacturers: expensive, bloated add-ons like HP’s iLo, IBM’s RSA and Dell’s DRAC. The DRAC isn’t too bad; at least it is capable of redirecting the BIOS boot screen down the serial line. But it’s still overly complex and involves messing around with java even just to log onto its web console. HP’s iLO doesn’t have the serial line redirect (although you can, fortunately, still access a console via ssh), requires an extra licence to make it do anything useful (seriously, who would be paying the large sums for an HP and not want those features?) and I have seen them crash all on their own more times than I care to remember. But my scorn today is reserved for the RSA card, truely the least pleasant of a bad bunch.
I’ve spent the morning trying to access the RSA console of a server in India that is not responding, but it’s an exercise in frustration. Initially, to set this machine up, a special driver had to be compiled and installed in the operating system for the console to even work, which is problematic enough – at least the cards from other manufacturers worked out of the box. Then the card had to be set to “Linux”, for reasons that escape me, before the keyboard would work. And now, I find that it only works under IE – under Windows, not wine. And there’s no other way of getting to the console, not even telnet.
Three years ago, when I was in Amsterdam, I was able to fix 90%25 of our out-of-hours problems with nothing more than a Nokia communicator. This is important, because it means that when you’re on-call, you’re not tied to your house and you don’t need to lug a laptop around with you everywhere. These days, a GSM phone couldn’t even cope with the bandwidth required to access the RSA card alone.
Sun had all this stuff right over fifteen years ago. Why do modern hardware companies find it so difficult to do?