Page MenuHomePhabricator

deploy elasticsearch/plugins to relforge1001-1002 servers
Closed, ResolvedPublic


Elasticsearch requires a few plugins in the context of Cirrus. Those plugins are at the moment deployed by trebuchet (migration to scap3 should happen at some point).

The relforge servers are in the labs-hosts1-a-eqiad subnet ( - 2620:0:861:117::/64). As such they do not have access to redis on tin, which is required for trebuchet deployment.

I'm not entirely sure how to move forward on this. I can't find documentation on a trebuchet / scap server for labs. It might be that relforge servers should actually be labs-support servers (and not labs-hosts). Or that we need to change the way we deploy those plugins for relforge.

I understand from the initial discussion on T131184#2251477 that those relforge servers should actually be in labs-support, but I am very much confused by all of this and might well be wrong. @chasemp, @yuvipanda, since you were on that initial discussion, could you help me clarify? Thanks!

Event Timeline

Aklapper renamed this task from deploy elasticsearc/plugins to relforge1001-1002 servers to deploy elasticsearch/plugins to relforge1001-1002 servers.Jul 23 2016, 11:42 AM

Some notes of our discussion with @chasemp:

Network categorization in hiera seems to be wrong. The subnet should be labs-support, not labs-hosts. DNS configuration should be the canonical resource and it does categorizes as labs-support1-a-eqiad.

Ferm rules on tin are generated from hiera (network.yaml), so correcting the categorization might just be enough to get all this working.

Change 302450 had a related patch set uploaded (by Gehel):
Labs network categorization correction

Change 302450 merged by Faidon Liambotis:
Labs network categorization correction

Correcting network categorization solved the plugin deployment issue. Closing this task.

renaming the network broke ferm on neon (icinga) -> T141957