Brando exception on Heroku

[rorygraves]

I saw there was a related Brando issue with Heroku deployment (chrisdinn/brando#64).
Is there any updates/fixes/workarounds known?

Thanks
Rory

[KarelCemus]

Do you mean this issue #29?

I could not reproduce it and as it is related to Brando, I haven't tried much. If you provide the minimal example to reproduce this issue, I might look at it.

[rorygraves]

Yes I do - I see this immediately on deployment with Heroku - would you be happy with a minimal example deployable to there?

[KarelCemus]

Sure. I'll look at it.

[rorygraves]

@KarelCemus This repo demonstrates the problem https://github.com/rorygraves/play-redis-heroku-issue

If I run against a local redis instance I see no problems (see readme for error and timing)

Against redisCloud on heroku it fails after 30 seconds -and errors until it runs out of memory and is killed.

[rorygraves]

My guess, from reviewing the Brando code is there is a race condition (its only a guess) between the command send, enqueing the senders and responses coming back from redis. Why this triggers on heroku but not locally is anyones guess ;)

[KarelCemus]

OK, I successfully reproduced the issue. I'll look at it. Thanks for your example

[KarelCemus]

Thanks for you example, I narrowed the issue down and reported as Brando #67. It turned out Brando does not properly authenticate against Redis. Plus there are several subsequent issues. Please see the issue.

[chandanpasunoori]

any progress on this issue..?

I'm not able to connect to redis

val redis = system.actorOf(Redis("localhost", 6379)) 
[INFO] [08/07/2016 10:32:28.971] [redis-test-akka.actor.default-dispatcher-3] [akka://redis-test/deadLetters] Message [akka.actor.PoisonPill$] from Actor[akka://redis-test/user/$c#-932188988] to Actor[akka://redis-test/deadLetters] was not delivered. [4] dead letters encountered. This logging can be turned off or adjusted with configuration settings 'akka.log-dead-letters' and 'akka.log-dead-letters-during-shutdown'.
redis: ActorRef = Actor[akka://redis-test/user/$c#-932188988]

[KarelCemus]

Actually, this is not an issue of play-redis, it is more likely issue of the Brando connector. I tried to push the PR forward but as you can see it is still pending.

Recently, I am considering to switch to some other Redis connector as Brando does not have active support. However, the switch takes some time and requires the library picking and I was quite busy. However, it is on my roadmap.

If you (any of you) have a recommendation to a good asynchronnous Redis connector library, I'd appreciate it.

Note: I might be wrong but your provided logs does not confirm this issue, these are logged regardless it. Are you sure this is the cause?

[chandanpasunoori]

yeah, I have seen Brando PR for this issue and updated artifact locally. and its working now

[chandanpasunoori]

I'm also in plan to update play-redis bando dependency to some other library.

[KarelCemus]

If you know about any other connector lib with active maintenance I'd appreciate it and switch the connector.

[chandanpasunoori]

https://github.com/etaty/rediscala

https://github.com/debasishg/scala-redis

https://github.com/scredis/scredis

these are few which are in my mind, and last one is active.

but need to evaluate

[chandanpasunoori]

I'm using Redis 3.2 and Scredis is only supporting upto 3.0.0

scredis/scredis#4

one PR is already there with this support but still not integrated

[KarelCemus]

Honestly, I don't emphasize the support of the latest version as the cache api supports quite basic commands. Furthermore, it is intended to be used more widely than just with the latest version.

However, thanks for these tips, I'll look at them

[KarelCemus]

Alright, Brando seems to be definitely inactive. I migrated play-redis to new connector library, the scredis implementation. It seems to be highly efficient and builds over Akka. Although there are not much commits recently, there are no major issues reported and it has about 200 starts (it is in two repos, old and new). The migration will be included in 1.3.0 version.

However, there is still much work to do to complete this milestone, as I would like to support collections in the next play-redis version. In consequence, I'm gonna publish 1.3.0-M1 version now. I would highly appreciate if you could test it on Heroku with AUTH command as well as in your regular environments and let me know how it works to evaluate the scredis connector. I am afraid of some internal bugs or performance issues.

For more details on migration see #64.

For more detail on 1.3.0 milestone see issues. This version besides the new connector adds several new commands.

Thanks to @etsrepo for hints on good connector lib.

[KarelCemus]

play-redis 1.3.0-M1 has been released. It is available in the maven repository and documentation is here.

I would appreciate if you could test it on Heroku and is some production applications.

[KarelCemus]

I tested it on Heroku with heroku-redis addon and it works smoothly. I consider this issue resolved in 1.3.0-M1 release.

[chandanpasunoori]

wow! thank you.
will test and update soon sorry I was inactive

[KarelCemus]

I encountered some issues with Redis Cloud addon, however I think it was due to wrong redis configuration. I was not able to connect to default database provided by this addon at REDISCLOUD_URL. I was not even able to find that default database in the addon configuration.

However, with Heroku-Redis and REDIS_URL it worked. And as this addon also has password-protected access, I think the authentication issue is resolved.

If you encounter any other issues on Heroku let me know, but I do not actively use Heroku so I might need help to reproduce the issue.

[chandanpasunoori]

I will test on Heroku, Azure, and Aws.
our production enviroment is on azure and aws

so far now dev environment is working.