Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Vince Offer]

Saw the blog post on TF2, figured I'd stop by and let you know the update is quite nice. 

[MedO]

We should be able to handle this kind of load better next time - most of the initial problems were because of bad server configuration (we hit the TCP socket limit), and after some tweaking the only problem still happening is that apache processes sometimes croak and take all requests queued to them to nirvana. I'm not sure why this happens, but it was only every 15 minutes or so and I might be able to work around it.

RAM and bandwidth use were in the green the whole time. I'm not sure about CPU because that's a bit hard to measure on a virtual server, but it might have been a bit high. The site was running smoothly though even in the first minutes as far as I can tell, IF you were able to get hold of a connection - so I think CPU wasn't a big issue either. I've sent a query to our hoster to see if we can get them to raise the TCP socket limit.

Edit: I changed the configuration in a way that might work around the "apache suddenly dies" problem... there's no good way to test it though

[Shadow the Past]

Valve should warn you ahead of time when they're going to advertise you. >_<

[NAGN]

we sorta emailed them

[Anyone]

Huh, so the team fortress 2 blog linked here?  Hopefully along side the influx of new players, we'll also get a few more servers to help deal with the increasing amount of active players.

[MedO]

Yes, this wasn't all that unexpected -- I actually would have thought the hit would be bigger. The problem is that we never had a load like this on the "new" server before, and I didn't have any experience with this kind of situation, so I didn't know how the server would react and which knobs would have to be tweaked to ensure everything keeps working smoothly. I'm learning this administratoring as I go along, and now I'm slightly smarter again

[Shadow the Past]

It might just be a steady wave of people, as it wasn't really a huge blog update or anything, so there wouldn't be 10,000 people trying to see the blog. Plus it was really subtle among the other crap that they talked about. (They don't really say what Gang Garrisson even is in the short sentence, so people would pass up looking at it.)

Well at least the server didn't tear a hole into another dimension.

[NAGN]

I think it's mostly subsided, most of the joins are from the initial posting of the blog (alot of people keep tabs on the blog for updates)

we did get a heavy influx of players/servers and the website did go down a bit, but I don't see that many new members or something, maybe they just didn't bother to post feedback and stuff?

[Starky]

How much damage dose the flare do?

[RareForumcrab]

How much damage dose the flare do?

somewhere around 20 on impact and 2 afterburn.

[sabermaster1.2]

Also, the infiltrator can no longer pick up the intelligence when cloaked, which i do not approve of.
 

[Sun Tzu]

Also, the infiltrator can no longer pick up the intelligence when cloaked, which i do not approve of.

That is a good thing. Uncloak if you want to capture it.

[CobaltBW]

On the bright side, it means a constructor can't cockblock you with his autogun on the intel anymore.

The downside being that mindlessly spyrushing the intel is somewhat harder now.

So, don't do the latter.

[Z3r05t4r]

If this hasn't been pointed out yet:

The demo's animation in the class selection screen. He does some kind of greet, but it's almost like a special one during the times of WW2. Only pointing out.

[MedO]

Some more updates on the server situation:

Host Europe replied that the only way to increase the socket limit would be if we switched to a bigger server package - which is not an option imo.
I've been looking at different options for limiting the number of sockets we need, but so far I didn't find anything that is both effective and simple.

Then there's the "apache bid farewell, cruel world" issue. Looking at older logs, this apparently happened regularly even on normal days, so one or two people every day have probably seen that ugly nginx 504 error page before now. The added load just caused it to come up more often and affect more requests. It would be neat to find out why it happens, but the workaround seems to be working, so you should only get a few seconds delay now where you would have got an error before.