The Linux Extremist

The secret police of Haitian dictator François Duvalier, known as the Tonton Macoutes, supposedly had a novel way of torturing their male victims. They would strip the prisoner, tie him to a chair, take a large nail and a hammer, and threaten to drive the spike through the unfortunate man’s scrotum unless he confessed to whatever trumped up charges they had in mind.

I remind myself of this unfortunate chapter in Carribean history whenever I get into a situation that feels hopelessly complicated or painful; the comparison always makes me feel better. My taxes can be raised, I can blow out a tyre on my car, the power can go out, I can even get injured, but thankfully, none of it is as bad as that.

In fairness, dealing with the fragility of Windows based solutions is not as bad as that, either. However, it’s bad enough that I am having to think about that to bolster my morale more often than I should.

I have dealt with the following scenario a number of times:

1. A server needs to be moved from one hosting centre to another.

2. The DNS entry on the server is changed so its Time To Live is set to zero.

3. The server is uninstalled, moved, installed in a new location and checked.

4. The DNS entry is updated.

5. The server is live again, and apart from minor hiccups, no one is the wiser.

I’ve worked through these steps with Solaris, FreeBSD, Linux and Windows servers; the only instances in which careful preparation has still yielded pretty awful results is with Windows. My most recent experience was involved just such a server continually crashing after a move of 100 miles from one data centre to another; a problem which had never manifested itself previously, suddenly appeared just by it having taken a quick trip down the road. Even with a group of Microsoft experts on the case, the best they could come up with upon examining the error logs was that some irrelevant files were missing. The files in question, the experts said, had somehow magically became relevant in transit. In the past few days, the server has improved, but none of the experts have taken the credit for the increase in performance, because they cannot see how they fixed it.

Much of the blame, in my opinion, should lie at the feet of the Microsoft philosophy of bundling everything into the OS; without tools being small and specialised, a complex series of dependencies is created. The operating system becomes a house of cards, and the tiniest shifting in a single element can cause its functioning to come crashing down.

Now I am in a familiar situation; I have to watch a moved Windows server like a hawk. My team of experts is monitoring it too, trying to find some rationality behind the server falling to pieces. At the moment, we’re still mystified. The only thing that is certain is that it will be an adventure, and any fix will have the feel of fragility about it; hitting upon an answer is just as much a game of chance as it is coming up with solid, educated guesses. Worse, if we have to make any more infrastructure changes, it is entirely possible that it will break yet again.

When Total Cost of Ownership calculations are made, it is rare that such maintenance costs are put into the mix; I laugh every time Microsoft claims a lower TCO than Linux because they always ignore real-world problems like this. Fortunately, I work for an organisation that is rational and businesslike enough to assess these issues. Not surprisingly, they have made a firm decision to move to Open Source at the earliest possible opportunity. I’m lucky; in a little while, I won’t need to think about the Tonton Macoutes to feel fortunate. For those who remain stuck with Windows’ House of Cards, maybe they don’t need something as extreme as that to make them feel better; but it’s unlikely that anything Microsoft serves up in the near future is going to make them feel good.

Trackback URI | Comments RSS