Opened 6 years ago

Last modified 6 years ago

#5969 new enhancement

Run the DP 2.0 in a working node

Reported by: vegasm Owned by: daniel
Priority: major Milestone: DP2.0
Component: DownscalingPortal Keywords:
Cc: vegasm, daniel


Assign a working node to the new Downscaling portal that runs outside Oceano. Oceano is overloaded several times and the Downscaling portal needs to be isolated.

Change History (2)

comment:1 Changed 6 years ago by vegasm

  • Component changed from TracMeteo to DownscalingPortal
  • Owner changed from antonio to daniel

comment:2 Changed 6 years ago by antonio

working nodes are in a IP private LAN.

This LAN is not accessible from the public web frontend ( which is in different building than the CPD where WN are.

This introduce an logic complexity which right now is been solved using oceano.macc as load balancer for apps.

WN (app backend) <-[ReverseProxy]-> OCEANO (loadbalancer) <-[ReverseProxy]-> METEO (web frontend)  

This requires 2 reverse proxies approach (OCEANO<->METEO) and (WN<->OCEANO).

This is reverse proxy cascade is possible but is not very well documented and explained. The AJP has some issues handling X-Forward HTTP headers and some manual "plumbing" is requiered.

The advantage of having this configuration is that we could deploy load balanced apps.

Currently TDS7 is deploying like this, but Tomcat instances are static.

I'm migrating to a dynamic cluster of instances of Tomcat. This will allow to adjust the apps demand.

To comply with the DP must be "portable", no having problems with multiple reverse proxies and robust of concurrent access to zones.

Note: See TracTickets for help on using tickets.