I have two Rails apps (one for web, and one for backend) accessing to the same PGSQL database. I would like to notify the other app if one app changes a table in the database.
How should I go about it?
4
-
maybe you could use thisjomo– jomo2013-09-15 09:06:25 +00:00Commented Sep 15, 2013 at 9:06
-
1What reason to do it? May be your design is wrong? I.e. if you want to use cache sweepers when some data changes, instead you can turn on using updated_at in cache keys, and so on.mikdiet– mikdiet2013-09-15 09:33:44 +00:00Commented Sep 15, 2013 at 9:33
-
Hi MikDiet: my two servers run independently, one for computation and the other for web. I think it's more than just cache sweeping.AdamNYC– AdamNYC2013-09-16 03:32:40 +00:00Commented Sep 16, 2013 at 3:32
-
@AdamNYC What solution did you end up implementing? I am also interested. In my case, a Java program writes to a database table, while a RoR is used pretty much as a GUI to generate reports.Martin Velez– Martin Velez2014-02-10 03:54:27 +00:00Commented Feb 10, 2014 at 3:54
Add a comment
|
2 Answers
I think this depends on:
- How reliable you need it to be.
- How fast you need the notifications to be delivered.
The FAYE solution suggested by @techvineet provides a good fast but unreliable option. (N.b. I don't mean it'll fail often, but it likely will occasionally, maybe 1/1000, If that causes you a problem, then avoid)
If you need something 100% reliable, and speed isn't important, you could write audit events to the database, and then poll that table from each app, if these are committed in the same transaction as the actual work is done, you should be safe... But it'll be as slow as your polling cycle.
Lastly, you if you want something fast AND reliable, then you could look at using something like ActiveMQ or RabbitMQ to give you reliable messaging between the applications to notify changes. You'll need a worker process in each app to listen to changes and deal with them appropriately.
My last comment would be that this 'smells' a little. The fact that you're trying to do this makes me think the architecture of your app might need looking at in the longer term. An obvious way of doing it might be to encapsulate all the business logic into an app which exposes an API, and then calling that API from both front and back end applications.
1 Comment
Richard_G
Hi Paul. Can we get some further explanation as to why the suggested solution is unreliable and why the 1/1000 failure estimate? Also, regarding the API solution, given that any such interface should be stateless, why would that solve this particular problem? Thanks...
You can try using Faye http://faye.jcoglan.com/, which is a publish and subscribe messaging server. It can be integrated with Rails https://github.com/jamesotron/faye-rails.git. Messages can be transferred from one app to another by subscribing to the messages and publishing.
Hope this will help.
