Doesn't play well with redis-namespace

[blarralde]

I couldn't quite figure out where it happens in the code, but somehow namespaces are not added to the main hash key. Example of what I see in redis:

redis = Redis::Namespace.new("sidekiq", redis: REDIS)
redis.keys
=> ["sidekiq:processes", "sidekiq:stat:failed", "sidekiq:sidekiq:status:b635df490af7a6fc0441892f", "uniquejobs", "sidekiq:uniquejobs:4a3b0ea5cd32400c837ec73565aa178d"]
redis.hgetall "uniquejobs"
=> {"ac11feaaea4378ba6f1f7df0"=>"sidekiq:uniquejobs:4a3b0ea5cd32400c837ec73565aa178d"}
redis.get "sidekiq:uniquejobs:4a3b0ea5cd32400c837ec73565aa178d"
=> "ac11feaaea4378ba6f1f7df0"

So somehow the hash doesn't get namespaced, but the unique job key inside of it does.

My guess would be that:

• the unique job key digest somehow receives the namespace in https://github.com/mhenrixon/sidekiq-unique-jobs/blob/11f88a3c5d5f7058d988f9ffd7854b1dfd2cea04/lib/sidekiq_unique_jobs/unique_args.rb
• the redis connection the gem is using isn't quite Sidekiq's

[dpetersen]

I don't know a lot about this gem, but I'm trying to sort out a problem with an app that uses it and I was trying to figure out if it was namespace safe. It appears not. As far as an explanation, I haven't had time to test this out, but the gem uses Lua scripts that are evaluated and run on your redis server. I'm guessing that it circumvents the way redis-namespace works. I think you can configure a unique_prefix (as seen here and used here), but it still doesn't appear to namespace the key, it just adds something to the way jobs get hashed so that different apps that work on a shared redis won't collide.

[mhenrixon]

Support for namespace has been dropped.

[aCandidMind]

Wouldn't it make sense to drop that code then?
https://github.com/mhenrixon/sidekiq-unique-jobs/search?q=namespace

[mhenrixon]

👍 @aCandidMind

[mengqing]

Any reason why support for namespace has been dropped? We run our sidekiq server on a namespace and currently this doesn't work (eg: when specifying replace when on conflict, the existing job doesn't get replaced)

[mhenrixon]

Partly because doing so means supporting a bad way of doing the Redis thing, partly because I am lazy.

• http://www.mikeperham.com/2015/09/24/storing-data-with-redis/