What is this running on?

meta
1

This instance of Discourse is currently running on Nginx 1.5.0 (dev) with Phusion Passenger 4.0.4.

Discourse itself is a Ruby application; Passenger is the application server for Ruby. Using Passenger lets me stick with Nginx for the heavy lifting. It also uses PostgreSQL for the back-end database, and Redis for caching and other stuff.

The pages you’re served are almost all generated via client-side javascript (check the page source!).

It’s definitely different from a traditional phpbb forum application.

2

And NOW we’re using Nginx 1.5.1 and Phusion Passenger 4.0.5. We’re also using Ruby 2.0.0-p195.

Feeling justified in splurging a bit when I was building this web server last year. Got PLENTY of headroom available—the thing that’s holding me back more than anything else is Comcast’s bandwidth. Wish they had more affordable business class plans that I could upgrade to. Don’t feel like paying another $40 a month just to get more upstream bandwidth.

3

For hardware, BTW, the CoG web server (which this is running on) is:

  • An Intel Core i7-2600S CPU @ 2.80GHz (32nm Sandy Bridge, low-power variant)
  • An Intel BOXDH67CLB3 LGA2011 motherboard
  • 16GB PC3-10666 DDR3 (1333MHz) RAM
  • One Corsair 90GB Force GT SATA III SSD

Cheap, and frankly, oversized even for what I’m doing (and a similar amount of hardware would cost $200 or more per month if I was paying for dedicated hosting).

6

Oh, wow, that’s awesome. For a full dedicated server, those prices are ridiculously good.

8

I did a quick update to make sure we were on the latest code revision. You might have been refreshing or logging on right when I was doing that.

9

HELLO TOPIC FROM TWO YEARS AGO.

Just FYI in case anyone is interested, we jettisoned Passenger some time ago. Discourse is now a containerized Docker application.

Our web stack is HAProxy (for SSL/TLS termination) -> Varnish (cache layer) -> Nginx 1.9.1 -> Discourse Docker (which contains Nginx and Unicorn and all the Ruby bullshit). Discourse also uses postgresql and redis.

10

I thought it was running on sunshine and dreams!! WEEP!! RUIN MY DREAMS MR. SMILING DINOSAUR WHO I USE TO KNOW AS KEEP AND HUMP THEIR LEG AS WELL, JACK THE INVISIBLE YORKIE!! ~flees into the shoe section`

11

I need some secobarbital in the shoe section please. Also clean up.

1 Like
12

I don’t know if the latest upgrade has anything to do with what I’m seeing…

Since I got back, these forums have been really, really slow to load. Looking at my network traffic, my browser is pulling down some insane amount of data from the server. On a mobile device (3G connection), it’s so slow as to be useless.

13

I’m seeing some complaints regarding the latest “updates” of Firefox and Chrome. Not sure if Opera, Safari, or IE are acting the same.

Oh, what am I saying? IE is always the browser of last resort - like strychnine is the painkiller of last resort.

14

Hmmm. Not sure wtf that is. Did a quick update to the latest codebase again just now, so let me know if it’s still going on.

15

No real testing done, but it seems faster today. If I remember, I’ll try to measure things.

16

FWIW, the built-in little timer tool thing that I see as the admin doesn’t show anything amiss:

If it starts to look weird again, pop open the Chrome dev console if you can and see what elements are lagging.

17

Running on - running on empty

Running on - running blind

Running on - running into the sun

But I’m running behind

1 Like
18

I’m running on fumes.

In this neighborhood, it’s mostly pot fumes, crack fumes, taqueria fumes,carbon monoxide fumes…

19

Nothing wrong with taqueria fumes.

1 Like
20

The UPS for the two web servers died yesterday out of the blue. No idea why. Tried replacing the battery, but it didn’t help—something in the actual UPS itself is broke. Still works in bypass, so I’ve got a new one ordered and the web servers are running without battery backup.

There’ll be some more downtime tomorrow when the new UPS shows up and I power the servers down to plug it in.

21

Did the magic smoke containment unit fail?

22

:flip: :flip: :flip:

No, it’s due to… Reverse bit-wise polarity mismatch in the I/O handler driver database.

23

Honestly I have no idea what went wrong. The three things connected to it (two web servers, one more server that runs prosody and bind9) died and the UPS started beeping like it’d lost mains power. Went into the closet and looked at the UPS and it showed 0% battery remaining. Moved the three servers over from the battery side to the bypass side and they turned right back on; did a bit of testing and none of the battery-backed outlets are providing power. Swapped out the UPS’s battery and got the same result.

So my official diagnosis is “it broke.”

It’s more than a year out of its 3 year warranty, so all I could do is shrug and order a new one.