Bluegreen deployment, old version stopped before new is promoted

Hi,

I have an app running (2 desired instances) with bluegreen deployment. Everytime I deploy, I expect it to upgrade to new version silently without disturbing any incoming requests (timeout or 5xx).

However, I noticed that there is short timeframe where incoming requests are not being routed to either the new or old instances (which lead to timeout/no reponse).

Any ideas why? Am I doing the wrong type of deployment?

This is flyctl status, new version instances are ready, passed health checks, waiting to be promoted, but old version is stopped already.

App
  Name     = service          
  Owner    = personal                
  Version  = 12                      
  Status   = running                 
  Hostname = service.fly.dev  

Deployment Status
  ID          = 57399ff7-bbaa-234e-d550-5b145f2d8b5f         
  Version     = v12                                          
  Status      = successful                                   
  Description = Deployment completed successfully            
  Instances   = 2 desired, 2 placed, 2 healthy, 0 unhealthy  

Instances
ID              PROCESS VERSION REGION  DESIRED STATUS  HEALTH CHECKS           RESTARTS        CREATED    
8deb9786        app     12 ⇡    sin     run     running 1 total, 1 passing      0               1m5s ago  
0f3b8f72        app     12 ⇡    sin     run     running 1 total, 1 passing      0               1m5s ago  
4da59036        app     11      sin     stop    running 1 total, 1 passing      0               7m8s ago  
d1fb1396        app     11      sin     stop    running 1 total, 1 passing      0               14m39s ago

This might be helpful: Internal network DNS propagation and new deployments - #4 by kurt

Will try, I don’t run nginx currently, it’s facing directly behind Cloudflare.

So scaling to 3+ should be ok?

Running 3 instances using bluegreen still experience some timeouts, might need to wait until official fix.

Hope to try flyio again soon!