Print

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

[Baughn]

The current spate of problems were due to running out of tcp buffer space, which apparently causes fastcgi (well, php-cgi) to puke.

I've switched back to plain old cgi, for the time being. It's nowhere near as fast (no "fast" in the name, see), but it at least *works*. Mostly.

But I have to agree, a dedicated server is looking increasingly tempting.

[Another]

I am getting
SYSTEM WARNING: session_start() [function.session-start]: open(/tmp/sess_ibfq3****************, O_RDWR) failed: Permission denied (13)
at the top of every bugtracker page.
Just prior to its appearance I failed to add a note to a bugreport with nginx error screen.

* - edited out as it may be a hash of some personal ID information.

Edit: It seems that now I cannot add any more notes to the bugtracker. I may need some admin help if it is only my account that is affected.

[Baughn]

An accident. Try now?

[Schilcote]

As far as dedicated servers go, you can usually just re-purpose any respectable computer and it'll work great. My SRCDS & "whatever needs to be computed right now" server cost me about $50 for a new hard drive, I found it in the garbage. You'd be surprised what people will throw out.

[Baughn]

I'm pretty sure co-location would be more trouble than it's worth.

[Another]

An accident. Try now?

All is working now. Thanks.

[Shades]

I'd seriously recommend looking at Varnish to run in front of nginx. It can't do anything about the post pages but all the page views (which is the vast majority) are going to be helped by it.

I've see an average of ~10 fold increase in handled 'get' pages from sites where I've done this myself. I've seen people claim more too but I can't vouch for that. I understand there is a move in progress that should solve a lot of this but if you find yourself reaching the limits of the new vps then it would be work looking into.

[Baughn]

"Amusingly", the current VPS doesn't have enough memory to do that, but nginx' built-in cache seems to be working.

The new one.. first off, I think it's more likely to be a dedicated server. That aside, yes, putting a cache in front could be a good thing to do.

Why should I prefer Varnish over the built-in cache in nginx, though?

[Shades]

I've heard good things about the nginx cache too but I've not had much experience with it so wouldn't want to recommend it. I also don't know what causes it to decide a request is cached (logged in users for example are always fun) where as I know you can tell varnish to cache parts of the page per user and parts globally to solve that.

Any memory heavy cache is good though, I only mention it as you mentioned it was hitting the php layer and fastcgi which shouldn't happen with a cache.

[Sensei]

All morning, it was "timing out" when I tried to post...

[cganya]

I actually haven't encountered any errors today at all.

[Tarran]

I haven't either.

[Xeriar]

I'm getting them more now too.  It's not part of the move process.  Generally the recent forum problems have been from traffic, but there's a little less traffic now, so the only thing I can think of is DF talk downloads?  In any case, the move will be initiated soon, but it looks like there will be some bumpiness first.

edit: what q said

Since nginx requires php to be running as a standalone process for fastcgi, it will throw gateway errors when all processes are unresponsive. More php threads usually will not help, it's usually caused by waiting for the database to respond.

[Urist McDepravity]

Since nginx requires php to be running as a standalone process for fastcgi, it will throw gateway errors when all processes are unresponsive. More php threads usually will not help, it's usually caused by waiting for the database to respond.

Not in case of VPS. These have limit on open file handlers/tcp connections/tcp buffers. When it hit the limit, it starts silently dropping connections between nginx and php process, which results to connection timeout on nginx side.

[Xeriar]

Not in case of VPS. These have limit on open file handlers/tcp connections/tcp buffers. When it hit the limit, it starts silently dropping connections between nginx and php process, which results to connection timeout on nginx side.

You're not using sockets for fastcgi? You should at least have enough file handles.

Regardless, hosts with overrestricted VPS settings are one of a number of reasons why I avoid VPSes like the plague.