Load Test the New Slashdot Setup

As many of you have noticed, Slashdot has been pretty bogged as of late, but the end of that is at hand. The new setup has a dedicated SQL box and 3 load balanced http servers. We've been spending the last few months getting everything ready, and now its time for the final check. I'd like to ask you help testing the load balancing stuff: You can access the new system at this address. Its a slightly out of date mirror of Slashdot, but it should be more than enough for us to determine if the new setup can cleanly handle the load. Thanks for you help: If this all goes well, we'll transition to the new setup in the next few days. (And just in time, Slashdot was kicking out like a thousand pages (pages, not hits) a minute earlier today and stuff was really bogging down)

Comments

CommentLimits?

[Trepidity]

So will this new setup mean we can finally get rid of those annoying CommentLimits of 25 comments per page?

Re:Put "10.1.1.1 adfu.blockstackers.com"in /etc/ho

[Signal 11]

Depending on your router tables, this may be a bad idea. If you use a default route and are on a ppp link (or forward it to another machine that does the same thing), it'll forward your query to 10.1.1.1 to your ISP, which may bounce all the way out to a major backbone before getting icmp unreach'd. In which case, it will spend another 4-5 hops getting back to you.. for a round trip of around 300ms.

The correct solution would be to set it to 127.0.0.*, which is guaranteed to have a fast turnaround. Depending on your setup, you can spit back a conn refused right away, or 404 when it accesses your local webserver.

--

Why not make slashdot better cacheable?

[Holger]

I was always wondering why slashdot is so heavily depending on dynamically generated pages. Even if just the main page had proper caching information, you could already hand off all the anonymous users (people not logged in) to the worldwide caching hierarchy out there.

See what the cacheability engine on ircache.net has to say about slashdot.org. At least the gifs are cacheable, but why is that "Expires"-Tag set? Do you expect the gifs to change every week?

Flooding moderation (re: LIme aNd c0c0 nutz!!!!)

[mcc]

perhaps this is just another way of "testing" slashdot? although he's testing the wrong server, dagnabbit. He's supposed to be flooding the _test_ server, to see how it stands up. Ha.

A note to Taco-san or other /. admin type people: maybe you should consider adding a special "flood" category to the Moderator list, along with Offtopic and Troll.. then set up the servers in such a way that anything with a "flood" rating would immediately be given a much much shorter amount of space to run before the cutoff for "Read the rest of this comment.." happens. That might help to cut down on scroll times when floods like this one happen, without hurting posts that just happen to be really long.

and God dammit, what the hell IS the name of that "put the lime in the coconut, drink a bowl up" song? that's a cool song but nobody seems to know where i can find it. :P

-mcc-baka
INTELLECTUAL PROPERTY IS THEFT

Re:Why is it...

[MindStalker]

Well its more than what you would think. Slashdot is tuned to the number of users it gets in a day. Latly even slashdot has been going down on rare occasions thats why they are getting new servers. The reason the slashdot effect is so much more hardhitting is the fact that if you take a site that say gets the same number of hits per day as slashdot. Then link to that site from slashdot, assuming you have no crossover (eg persons who would have visited the site anyhow) you have instantly nearly doubled the amount of visits to that site for a short amount of time. This sudden doubling effect generally leads to a very buggy site which causes people to hit the reload button. This increases the page hits by several percent more. Finally bringing the server to its preverbral knees