Getting Discourse installed on OVH VPS

It’s not been long and they recently came out of beta IIRC. Their EU network is excellent and North American network is decent, but you’re better off with DO if your traffic mostly comes from elsewhere.

Traditional Intel perform much better for this type of application, but power8 is interesting as well.

I am going to run a fairly large discourse community in near future and I’m most likely to go with OVH’s http://www.soyoustart.com/us/offers/e3-ssd-1.xml server. Later, if it gets more popular, I can just move web components out of it and run them in on-demand runabove instances. Object storage is dirt cheap as well as bandwidth, so it’s an ideal setup for me. Moreover my traffic is mostly European, so it’s the most balanced setup for me ATM.

Kimsufis are fine for exploration, but never run a production in it. Soyoustart exists for a reason. :smile:

1 Like

I already got an instance of discourse up with runabove :smiley: I choose the dokku install and went straight to installing discourse. It comes installed with nginx, so quickly shutting down the service and no worries!

The soyoustart is a good deal, and I’d go with that for a larger discourse community, but the kimsufi is still an amazing deal for a dedicated that would work for a medium sized community without issues.

I have 32gb on a xeon e-1220 and 250gb samsung ssd. I don’t see discourse needing more resources than that. The only limitation I’ll run into is bandwidth. I have uverse with 45/6 package, but will be upgrading to gigapower as soon as it’s available. I’m in one of the early cities to be implementing it. As long as I keep the radio station broadcasting, and podcasts with OVH, there shouldn’t be any issues. Right now the discourse is allotted 3 cores and 3 gigs, but can be bumped up as needed.

I’m curious, what would you consider a “large” discourse community? We are expecting a few hundred members by the end of February. It could get over 1,000 by the end of the year, but I’m not sure. I’m wondering how many resources would be needed per lot of active users. It seems the best answer would be “it depends” , but rough figures would be nice to get an idea.

Great. Glad to hear your could get up and running without any hassle.

It’s a good deal with reasonable redundancy. My only problem with kimsufi is single disk setup and if you experience any hardware failure, you’re doomed. For example, if your disk dies Friday night, they won’t replace it until Monday, which would be big problem for me. At least with 3 disks, I can have some redundancy with soyoustart server, and it also gets faster hardware replacement.

Your setup shouldn’t be any issue IMO, that’s plenty of power to keep it happy.

I haven’t performed any benchmarks as we’re still developing, so I can’t give you exact figures. But a decent sized VPS from providers like DO or runabove should be able to handle at least 50req/sec judging from http://www.isrubyfastyet.com/ so I don’t think you have anything to worry about.

We’re expecting around 2k users in couple of months of launch and we may break 10k mark by June depending on how things go. I think http://www.soyoustart.com/us/offers/e3-ssd-1.xml shouldn’t have any problem handling that. But I think the real defining metric should be concurrent users rather than total users. I’ll post the benchmark results after we finish development.

1 Like

Keep in mind that’s with a mere 2.2ghz c2d. Our xeons should be at least 2.5x as fast per core. This is single threaded performance as well. If everything scales linearly, I’m guessing that means at least 1,200 requests per second, which means a 10k member community should not be an issue. Now, how this actually pans out in the real world, we don’t yet know.

Indeed. We’ll know more after more testing. But keep in mind that benchmark was probably minimal rails app, without database server running in same machine. But I’m not worried about computing power or memory at this moment. I can always throw in faster hardware and still run the site at reasonably low cost because servers are dirt cheap. I can always use a dedicated server for database and move everything to runabove instances and still pay less than $150/mo.

1 Like

Servers are quite cheap these years. In all honesty, the majority of the tech improvements are best utilized for military/aerospace applications. We petty consumers are spoiled beyond belief, and without the capabilities to effectively utilize all of what’s around us.

1 Like

I’ve created a guide which includes my solution here:

My main err was assuming bind9 was of absoute importance. It wasn’t. Bind9 is necessary if you would like to host your own nameservers and run a control panel. Once I tried everything raw, the process went smoothly. I allowed godaddy to host the nameservers, created A records for each doman/sub pointing to cloud IP, then setup nginx reverse proxy on the front-end with discourse and apache vhosts on the back end. All is well with OVH when you do away with control panels, nameservers, and a need for bind9.

Give me a little slack, I didn’t even know what ssh was 90 days ago, LOL!

2 Likes