Fix: Shutdown hub when transport is gone

[bpolaszek]

Fix #24 - Hub is not aware of redis connection losses

• Introduces a regular ping to Redis connection
• Whenever a ping fails, kill the hub

Unfortunately I was not able to set up an automatic reconnection.
Updates can be prevented from being published by returning a 503 error, but for subscriptions, I could not manage to :

• Stop the SSE
• Make it aware of a new redis connection

Killing the hub is the best I can do so far - a supervisor process (systemd, supervisor etc) should handle its respawn.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (6fc70c7) 100.00% compared to head (512df63) 100.00%.

Additional details and impacted files

@@             Coverage Diff             @@
##                main       #25   +/-   ##
===========================================
  Coverage     100.00%   100.00%           
- Complexity       138       141    +3     
===========================================
  Files             25        25           
  Lines            396       409   +13     
===========================================
+ Hits             396       409   +13     

Files	Coverage Δ
src/Hub/Controller/PublishController.php	100.00% <100.00%> (ø)
src/Hub/Hub.php	100.00% <100.00%> (ø)
src/Hub/Transport/Redis/RedisTransport.php	100.00% <100.00%> (ø)
src/Hub/Transport/Redis/RedisTransportFactory.php	100.00% <100.00%> (ø)
src/functions.php	100.00% <100.00%> (ø)

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[bpolaszek]

I just added a commit - can you please try again with this fix?
I assume it's already done but don't forget to configure pingInterval in your DSN for this to work.

[misaert]

I think I don't need to configure pingInterval because it is defined to 2 seconds by default, right?

The new commit seems to fix the issue 👍

But I think the changes on src/Hub/Controller/PublishController.php is unnecessary. WDYT?

[bpolaszek]

I think I don't need to configure pingInterval because it is defined to 2 seconds by default, right?

The new commit seems to fix the issue 👍

But I think the changes on src/Hub/Controller/PublishController.php is unnecessary. WDYT?

Good news! 🎉
IMO throwing an exception when publishing fails adds an extra layer of reliability as it does not require to wait for the ping to fail (if configured) to throw an error. Besides, this ensures that the hub does not respond with a 201 status when publishing actually failed. WDYT?