f13.net

This is not a rant.

Back in the early days, weekly Maintenance took four hours.

Then they extended it to six. That became the new norm.

Now they've extended it to eight (http://forums.worldofwarcraft.com/thread.aspx?fn=wow-general&t=5791993&p=1&tmp=1#post5791993), by starting it earlier.

What's the deal? Are the databases just getting too overrun with requests requiring that much more time to fulfill? I'm no Oracle guru, and heck, don't even know if that's what they're using, but based on our old SWG discussions about that game's perennial DB problems, I'm wondering if WoW suffers a similar problem.

Anyone have any insights/thoughts?

Blizzard does not know how to run a service. They can make a helluva product, but a service is a different animal.

They also have added how many more servers in that timeframe?  They're taking 'em down and doing whatever to each one, then have to make sure it restarts.  That's going to add time to the window, because lord knows you can't just go hiring more people to keep it the same time. What would they do during the rest of the week.

Maybe they could be used to look at improving the database when they aren't fixing pressing problems?  God knows blizzard has the money.

As for the adding asstons of servers since release (they have) why not just stagger the maintenance schedule to provide less downtime to paying subscribers? Since there isn't any cross-server pollination of any kind, why not do half the servers on Tuesday and half on Wednesday. You still have 8 hours of downtime, but for the players, it appears to be only 4 because their server isn't being taken down for more than 4 hours. Sure, it might be inconvenient for Blizzard, but not for the consumer, which is what service-oriented organizations think of first.

Because junkies would flock from whatever server was down to whichever was up.  So basically your Main has more lag, but you can play an alt during maintenance.

Not worth it imo.

Vivendi Universal has the money.

This is a small but essential difference.

--GF

Nothing like logging on to find IF stacked to the rafters with 1,000 characters named after variations of "Bliizzardsuxlol," all whining because their home server Bleeding Asshole is down.

What's killing me are the Monday nights.  I think we're coming up on six weeks now that every Monday is unplayable.  After last night we decided to try and find another night to all get together.  If that doesn't work a few of our group may bail.  Sigh.

Yea, that's one reason I wondered Poco. Every Monday's a crapshoot on the login server for us. Ingame it's fine, but woe be those who LD.

As to staggered maintenance I imagine it's because they don't want to be running different version numbers concurrently. It seems already hard enough keeping 400+ servers going (I think that's the U.S and Europe totals, since The9 does the rest in China and other companies handle Taiwan, Korea, Malaysia, etc.)

Are you sure the downtime is expanding on a permanent basis? Usually it isnt that long, and occasionally they extend it to deal with special problems, or roll out a new feature.

You sure it isnt just the occasional spike in downtime length you are seeing, or is it really increasing and staying that way?

Yea, I'm projecting really. It was 4 hours and then 6 became the norm. I'm wondering if 8 will become the new norm. Then we got hit with a 2-hour maintenance today too.

I'm only wondering on the tech backend that requires this much maintenance. Is it due entirely to how many servers they have to push updates to? Is it something about clearing the database queues?

I have to support a large installation and I think the answer could be anything.  Manpower, the specific problem, or even just people getting tired of having to do six hours of work in four hours.  They aren't going to tell you what they are really doing, unless you corner one at a convention, and even then it might not even be discussed.  I can start talking about the stuff I do and watch the eyes glaze over, so I don't.

What exactly is "server maintenance" anyway? Normalizing a database? The most maintenance most servers need is to be stopped and restarted. (Actually, most servers don't need any maintenance)

The hamsters need to rest every once in a while. Shutting down all the servers makes it easier to backup the databases as well.

You'd have to be quite the junkie to start an extra "server maintnance char", I don't think it would be such a large problem.

If no one was applying patches to Oracle or SAP, we would not have much maintenance at all.  Even then, we don't reboot the suckers unless forced to for some reason... meaning critical bug or security fix that requires it.  So I don't get it unless they are doing lots of database work.  Our worldwide employees experience better than 98% uptime, unless they work in someplace extra shitty like Lagos, Nigeria or Pune, India.

It could be that they are doing offline backups in there from time to time.

You sound like the sort of person who'd know that they don't need to be doing this, were they not making a trivial error in their server spending.