If anyone comes across anything odd or has a problem with something, please let us know here!
pixls.us has been either unreachable or very slow lately. Whazzup? (of course, itās now a bit better otherwise I wouldnāt be posting this:))
Iām glad you posted this, as I meant to come in and point it outā¦
Ok, so thereās two separate servers involved in running the two sites, pixls.us and discuss.
There are also two separate issues right now with each of the servers. 
pixls.us
This is shared webhosting with stablehost here in the U.S. (for me). (Arizona if Iām not mistaken.) This hosting costs about $6/month (USD).
The Problem with the main site right now is that Iāve made it use a secure connection everywhere (https), and just recently apparently the host had implemented firewall rules limiting the number of simultaneous SSL connections. This means that various parts of the main website were refusing connections for assets (images, css, etc) randomly as that limit was reached.
This took a couple of days to figure out with their tech support folks, but once it was escalated they immediately identified the problem. We are now trying to find a solution. Theyāve increased the number of simultaneous connections at the moment, and it seems to have alleviated the problem on the main site for now. (It gets even stranger when I point out that it seems to only be a problem on webkit based browsers(?!) - possibly something to do with Keep-Alive in the headers not working).
discuss.pixls.us
The forums.
This is an instance of discourse that is hosted on a virtual server at digital ocean as it was recommended by many in that community. This server costs about $20/month (USD).
The problem is two-fold according to @darix, who set everything up for us here (and keeps earning more and more gift beer when we get a chance to hang out in person for all his hard work.).
-
RAM on the server. The current instance has 2GB, which is a bit more than what is recommended for a small-ish forum. Even so, darix feels that it could be increased (no doubt it certainly canāt hurt). This RAM limit may be a reason for some sluggishness.
-
Ruby-something-passenger-something. I must admit to my ignorance on this part of things, but what I can piece together is that Phusion Passenger, which handles requests to the server and queues things up as needed, got hung up when reaching 100 requests in the queue that went unserved. I donāt know what the solution would be at the moment, so I am occasionally restarting that service on the server which seems to get things back to normal-ish. (I did this last night, and is likely the reason things feel better right now).
Solution?
I am working with the shared host to mitigate problems on the main site right now. Hopefully weāll have that sorted out today.
I may set up a cronjob to restart the passenger service every X hours to keep the forum feeling snappy (until we can find a more permanent solution).
With that said, one of my options to move forward is to move the entire main site to the same server as the forums, and to move all of that to a new virtual private server (VPS) provider. This is still an option, and thereās a good deal on a server at hetzner.de, as well as possibly using a new server with one of the current folks we are using. Stablehost has a 4GB VPS for $30/month, and DigitalOcean has one for $40/month.
So we changed the forum hosting to follow a bit more closely what upstream normally does (-passenger -apache2 +unicorn +nginx). Letās see if that helps.
2 Likes