You have to configure a few things: - determine set of names to configure public cname, hosts file entries or local dns cache entries, and virtualhost confis on a proxy. ex: supervisor01.mycompany.com. - reverse proxy front end, like nginx with routes back to appropriate backends based on virtualhost config - manage /etc/hosts entries for each node (or local dns cache) - `hostname -f` must match appropriate name, ex: supervisor01.mycompany.com - public dns entries (CNAME) that all point to the reverse proxy server. - if aws, you can config elb to point to reverse proxy end point for security. - also aws alb might work as front end reverse proxy end point, have not tried it personally.
I tried this on aws w nginx running on one of my zookeeper servers. works well. I have some shell scripts online that demo how to set this up: - nginx config lib - https://github.com/darkn3rd/storm-vagrant/blob/master/scripts/lib/nginx.src - how to use lib in setup_proxy(): https://github.com/darkn3rd/storm-vagrant/blob/master/scripts/lib/vagrant_driver.src On Jan 22, 2017 1:32 PM, "Alexandre Vermeerbergen" <[email protected]> wrote: > Hello, > > I deployed our application to an Apache Storm cluster (currently version > 1.0.1) on Amazon E2C instances. > > Nimbus UI works like a charm, except for the "debug" links which URLs are > based on private FQDN instead of public FDQN address of VMs. > > For example, when clicking on the debug link of one of the executors, I > have a URL looking like this one, which isn't accessible from outside the > VM: > > http://ip-161-12-410-48.eu-west-1.compute.internal:8000/dumps/<name of my > topology> > > Please note that the "Host" on the same line is > ip-161-12-410-48.eu-west-1.compute.internal, > so I guess that's the root of my issue ; I guess it should rather be a > public FQDN like ec2-12-132-11-1.eu-west-1.compute.amazonaws.com > > Any idea how to make Storm / Nimbus / Nimbus UI use public FQDNs instead > of private ones? > > Thanks, > Alexandre > > >
