Difference between revisions of "Good Practice"

From Open Rail Data Wiki
Jump to navigation Jump to search
m (Added category)
m (Add a recommendation to re-queue messages locally)
Line 3: Line 3:
 
* '''Sign up with one account''' - it gives an accurate reflection of how many people are actually using the service
 
* '''Sign up with one account''' - it gives an accurate reflection of how many people are actually using the service
 
* '''Connect to the service once''' - you can subscribe to multiple topics or queues in a single connection.  Don't use separate connections for each of the feeds
 
* '''Connect to the service once''' - you can subscribe to multiple topics or queues in a single connection.  Don't use separate connections for each of the feeds
 +
* '''Process messages quickly''' - messages left on the server build up and can cause problems for other users, so consider using an  [[Advanced_Uses#Bridging_Topics|configure an ActiveMQ bridge]] to keep the messages under your control
 
* '''Don't leave a failed client running''' - if you can connect but get an authentication or authorization error - '''stop your client''' - don't leave it continually trying
 
* '''Don't leave a failed client running''' - if you can connect but get an authentication or authorization error - '''stop your client''' - don't leave it continually trying
 
* '''Make sure your client handles failures''' - if your client fails to connect or is disconnected, use [http://en.wikipedia.org/wiki/Exponential_backoff an exponential backoff] and wait 1s, 2s, 4s, 8s, 16s etc. before trying to reconnect.  Service problems sometimes occur, and if everyone tries to reconnect every second, the service will not recover quickly
 
* '''Make sure your client handles failures''' - if your client fails to connect or is disconnected, use [http://en.wikipedia.org/wiki/Exponential_backoff an exponential backoff] and wait 1s, 2s, 4s, 8s, 16s etc. before trying to reconnect.  Service problems sometimes occur, and if everyone tries to reconnect every second, the service will not recover quickly

Revision as of 17:15, 17 October 2016

Please bear in mind the following points when you use the service - they're here to make sure everyone gets a good service:

  • Sign up with one account - it gives an accurate reflection of how many people are actually using the service
  • Connect to the service once - you can subscribe to multiple topics or queues in a single connection. Don't use separate connections for each of the feeds
  • Process messages quickly - messages left on the server build up and can cause problems for other users, so consider using an configure an ActiveMQ bridge to keep the messages under your control
  • Don't leave a failed client running - if you can connect but get an authentication or authorization error - stop your client - don't leave it continually trying
  • Make sure your client handles failures - if your client fails to connect or is disconnected, use an exponential backoff and wait 1s, 2s, 4s, 8s, 16s etc. before trying to reconnect. Service problems sometimes occur, and if everyone tries to reconnect every second, the service will not recover quickly
  • Use a durable subscriber for topics - if you want data to be queued in case you disconnect, request a durable subscription for the Network Rail platform when you connect
  • Use Stomp heartbeats to detect network problems that may otherwise leave your client hanging with no data