#metabrainz

where might I find instructions for setting up the search server? They aren't included in the musicbrainz-server repo and the search-server repo seems to be for the old search code...

agentsim: it depends on how you install(ed) musicbrainz-server: from source? docker? virtualbox?

from source, following INSTALL.md

then you can either install the search server (mb-solr) from source or as a docker container

maybe I'm being a little thick. The instructions in mb-solr don't seem to explain how to integrate with musicbrainz-server except to reference an old branch of code

and the docker instructions, "run docker-compose up" sounds great, but surely I'd need to configure what musicbrainz-server to connect to, wouldn't it need to configure auto-updates of the search index, etc...

metabrainz/picard#4995 (master - 9dab7d1 : Philipp Wolfer): The build passed.

Change view : https://github.com/metabrainz/picard/compare/02...

Build details : https://travis-ci.org/metabrainz/picard/builds/...

agentsim: musicbrainz-docker indeed assumes musicbrainz-server is in a docker container

the search-server repo is legacy

yvanzo: I get that search-server is legacy, I'm just not clear on how to run the replacement, mb-solr. It sounds like I need to go learn how to docker...

https://github.com/metabrainz/mb-solr has some instructions on how to install it from source

isn't that the search server with the broken cd stub search? ;p

agentsim: learning docker seems a better idea

i heard docker is easy

it is sensible to run both the main server and the search server as docker containers on one machine in a production env?

agentsim: because there is not just mb-solr to be installed, there is also metabrainz/sir and rabbitmq (plus sql triggers)

docker is sounding pretty good then

agentsim: I suggest you to give a try to https://github.com/metabrainz/musicbrainz-docke...

(the master branch is still using the old search-server)

yvanzo: thanks that looks very helpful. The CPU requirements are steep :)

yes :( this is mostly to allow rebuilding search index from scratch

plan is to avoid this by allowing to load search index dumps instead

so once the index is built, updating it doesn't eat as much CPU?

yes

that's not too bad then, I can launch with a 4 or 8 core, be patient on the initial index and then I should be ok?

yes if you can be patient for two days.

yep, shouldn't be a problem

note: the search index rebuilder (sir) is not self-adapting to the machine computing capacities, thus you need to slow it down via its config to avoid "thrashing"

good to know, thx