1

I created a chat module for my RoR system with private_pub gem (that uses Faye) and it was working very well until some weeks ago.

We tested and installed the module on Amazon AWS and it was in production mode with the rest of the application and suddenly we received some complaints from our customers reporting long delays when sending or receiving messages.

This was not like that some time ago. It took 1 or 2 seconds max to send a message (actually I believe much less than that). This machine on Amazon was never rebooted and is running the application (not the chat, the chat was installed around 4 months ago) for over an year now.

I restarted the Faye server (I am using TMUX to run the server and detach it from terminal and get it running on background) but it didn't seem to help.

Any ideas on what could be causing this behavior or how can I test it?

Thanks

Victor Ferreira
  • 6,151
  • 13
  • 64
  • 120
  • Do you have any CloudWatch of other monitoring going on? Can you inspect the logs? – John Feltz Sep 20 '16 at 16:51
  • we are watching Rails logs and nothing related to the problem is happening. nothing related to the chat, only some missing routes from time to time. and we are exhaustively testing the feature. takes from 5~12 seconds to send a message. and one of the roles (operator-client) is taking a little longer than the other. it's weird because the user with the other role is updating the database (time when last message was sent), the operator is not – Victor Ferreira Sep 20 '16 at 17:10
  • when working with a production you'll occasionally need to look outside your application trace for clues. Check your apps cumuative bandwidth, CPU, and database consumption. If it gets too large you'll get difficult-to-debug errors, and you'd need to refactor (improve app performance) or upgrade your cloud hardware / limits. It also might be worth setting up a staging server that you can spam with fake data to try and reproduce the problem. – max pleaner Sep 21 '16 at 00:00

0 Answers0