storm-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Joaquin Menchaca <jmench...@gobalto.com>
Subject Storm Cluster End-Points vs. Direct Access
Date Fri, 07 Oct 2016 05:15:44 GMT
Two things I wish were possible.

1) storm client

I wish that a storm client did not have to talk to every node in a cluster
directly, just interfaced with one end point, and that service fetched
everything as necessary.  I am not sure if all the systems that one talks
to is fully documented.

2) web services

For the web services (ui, logviewer), I wish we could specify routes.
Using host based routing through a reverse proxy is difficult, when routes
are more flexibly, e.g. /log/supervisor01, /log/supervisor02, /ui/nimbus01,
or whatever makes sense.

This way we can have one end point (a reverse proxy) that backends to the
appropriate systems.  Also, with containers, such as a supervisor in one
container, and a nimbus on another container, that live on the same system,
if they have separate logs, then you have to have two reverse-proxies, one
for the container routes, one for the base systems, or something else
creative.  Not being able to control how the URLs are rendered is painful.

-- 

是故勝兵先勝而後求戰,敗兵先戰而後求勝。

Mime
View raw message