MOVED error in Redis Storage config & Elasticache

[danDanV1]

Is lua-resty-auto-ssl compatible with aws redis elasticache?

It's all working fine with local storage, trying to trouble shoot getting a redis elasticache cluster working. Connecting with the cluster Configuration Endpoint.

        auto_ssl:set("redis", {
          host = "clustercfg.d1rp7blahblah.vqmcgo.use1.cache.amazonaws.com",
          port = 6379
        })

2018/04/29 04:28:56 [error] 23176#23176: *20 [lua] ssl_certificate.lua:134: get_cert_der(): auto-ssl: error fetching certificate from storage for www.domain.me: MOVED 15012 10.13.41.79:6379, context: ssl_certificate_by_lua*, client: 10.33.22.89, server: 0.0.0.0:4431
2018/04/29 04:28:56 [error] 23176#23176: *7 [lua] ssl_certificate.lua:134: get_cert_der(): auto-ssl: error fetching certificate from storage for www.domain.me: MOVED 15012 10.13.41.79:6379, context: ssl_certificate_by_lua*, client: 10.13.22.89, server: 0.0.0.0:4431
2018/04/29 04:28:56 [error] 23176#23176: *20 [lua] ssl_certificate.lua:77: issue_cert(): auto-ssl: failed to obtain lock: MOVED 10592 10.33.41.79:6379, context: ssl_certificate_by_lua*, client: 10.13.22.89, server: 0.0.0.0:4431

[danDanV1]

$ redis-cli -h blahblah.clustercfg.use1.cache.amazonaws.com -p 6379
blahblah.clustercfg.use1.cache.amazonaws.com:6379> SET foo 100
(error) MOVED 16008 10.13.01.79:6379
blahblah.clustercfg.use1.cache.amazonaws.com:6379> GET foo
(error) MOVED 16008 10.13.01.79:6379

Using the -c param on redis-cli command line client sorts it out. It follows the MOVED redirect from the cluster.

$ redis-cli -h blahblah.clustercfg.use1.cache.amazonaws.com -p 6379 -c
blahblah.clustercfg.use1.cache.amazonaws.com:6379> SET foo 200
-> Redirected to slot [12182] located at 10.33.41.79:6379
OK
10.13.01.79:6379> GET foo
"200"
10.13.01.79:6379>

Can we configure resty-auto-ssl to connect with "-c"? or perhaps detect the MOVED response and reconnect to the appropriate server?

[marcfowler]

Did you have any luck resolving this? I wasn't even able to connect using a direct shard endpoint (blahblah-001.xyz.0001.use1.cache.amazonaws.com). I don't have much experience with Redis and not entirely sure how to resolve this!

Edit: Looking back over, I assume this is because Elasticache is a Redis cluster and #39 confirms this isn't supported, though openresty/lua-resty-redis#102 (comment) points to a client that does handle it so perhaps that could be used instead?

Is there a way to use Redis without a cluster in AWS without manually running it via EC2 in the interim?

[danDanV1]

I'm looking into using this storage adapter: https://github.com/steve0511/resty-redis-cluster

Would be great if lua-resty-auto-ssl supported it natively.

Another quick solution is connecting directly to the node that is the primary read/write. Could do some hackish work arounds to detect which one it is and keep it up-to-date. Might be able to handle the MOVED response directly in the lua scripts with the current adapter?? I think using the redis-cluster adapter is more elegant. However, I'm in a bit of the same boat, no experience programming in Lua.

[jordanade]

I wonder if a solution has been found for this problem, or a workaround configuration for elasticache? Will it work if I create a Redis store with Number of replicas=0?

Or...what's the best workaround that avoids Redis entirely? Shared EFS or EBS mount?

[alexphelps]

Any updates on this? I've stumbled into the issue of needing a way to support a redis cluster storage backend as well.

[alexphelps]

I was able to adapt the existing redis storage adapter to work with resty-redis-cluster. Seems to work on my local, haven't used it out in the wild or with aws elasticache redis clusters yet.