I switched the forum software in August of 2014, and I upgraded all server hardware over the past 12 months. I even moved everything over to a new datacenter a few months ago.
They are all different because they do different things. For example, the database server is a quad core setup with 16GB of RAM, whereas the search cluster is a dual CPU, high-memory setup (40GB) because it holds everything in memory so results can be delivered quickly. The web server is dual CPU, with moderate RAM (8GB) and high storage to accommodate things like file attachments.
Sounds like the database server is a single point of failure, a common issue for most configurations. What is your recovery plan if you lose the database server for any reason? Just curious because I encounter this question regularly when planning for high availability.
Hi, Do you actually use all that hardware just for this site? Or did you over build it? I'm wondering because I would have thought the whole thing could run on a run of the mill dual core 2gb system. Do you have extra stuff running besides the forum or something? Or is there a lot more traffic on here than I realize?
Frequent backups. I just find that it's way easier and practical to frequently backup and restore to a fresh setup than it is to constantly be managing (and paying for) a bunch of redundancy that you rarely, if ever, need. I mean, I love ET as we all do, but let's face it, what we do here isn't exactly "mission critical". Many years ago I had a big issue that kept the site offline for almost two days. And you know what happened? Nobody was mad or upset and no advertisers left. I apologized anyway of course but ultimately everybody was just relieved that "ET was back up". In total, I think we've had about 3 days of downtime in 19 years. And the only reason we had that 3rd day was because a garbage truck somehow severed a fiber optic cable in the city where my data center was located. But hey, as we all know.... shit happens no matter what!
It's over built a little bit but we definitely use it all. It's the uploading of file attachments like charts and photos, and extensive searches that would bog down a dual core system like you described. Add some peak traffic to that scenario and the whole setup would come to a screeching halt pretty fast. I remember back several years ago when there was that big scandal about the NY Attorney General Elliott Spitzer paying money to some woman so he could get laid. I can't remember her name, but all I know is that somebody here on ET posted a bunch of photos of her before any other media outlet did, and Google instantly indexed it. So guess what was the first search result whenever anybody typed his or her name in Google after the scandal broke out all over mainstream media? The traffic buried us to say the least. And it went on like that for quite a while afterwards.