Brett Neumeier+wiki
Puma and Nginx production stack
Hi Brett,
I'd definitely agree with your assessment of Wagn's discoverability; it sucks. But I think we're going to make major strides in the next few releases. We've finally finished a bunch of low-level harmonization that will make it much easier to make progress on that front.
Definitely interested to hear how your deployment goes. I personally prefer postgresql, too, but we shifted to focusing more on mysql when we started supporting Cloudstore hosting. I don't have much experience with nginx / puma / unicorn but would be grateful to you if you shared what you learn in your experiments, even (especially?) if there are problems.
As for your revision issue, that's not something I've seen. What kind of caching are you using?
caching -- initially I used the filesystem store, but after the first time I noticed the behavior I switched to dalli_store. I still have noticed that sometimes when I edit a card several times in a short amount of time, an old version of the card shows up until I clear the cache; but that hasn't happened often enough to get a handle on what's going on (and it hasn't happened at a time when it was convenient to look at the log file and see if I can figure out what's up).
The problem has not recurred since I upgraded from the first version I set up (1.11.1 or 1.11.2, I don't recall which) to 1.11.3. I also switched from unicorn to puma at the same time... so, not very good science, I don't know what change eliminated the problem, but from a pragmatic perspective the problem is gone and I'm not all that concerned about exactly why. I notice that now the log file shows a lot of cache read and write activity, which it didn't do originally -- perhaps the original issue was a cache coherency heisenbug that is corrected by adding the logging activity? But I don't really have the energy to try to track that down.
Added documentation on a simple nginx+puma setup elsewhere, with the hope that someone will find it interesting. Later on I'll add notes on how I have my actual server set up -- I wanted to run multiple distinct wagns, with a single copy of wagn powering them all so that a change in one place becomes manifest in all the wagns after the next restart, so that's what I have in place. And also experimentation and performance testing for different ruby versions: MRI, rubinius, jruby.
Well, maybe not a well controlled experiment, but interesting and valid data nonetheless. Thanks.
curse you, heisenburg :) thanks for all this, Brett!