19:47:12 GMT <prsn> hi! does anyone know if there is a performance difference between using many top-level keys and using one hash with many entries?
19:47:21 GMT <prsn> this is in an AWS elasticache cluster
19:48:21 GMT <minus> prsn: hashes are more memory efficient, dunno about performance though
19:48:43 GMT <minus> prsn: redis-benchmark will tell you.
19:48:51 GMT <prsn> i read in the docs that very small hashes have a special memory-efficient implementation, are you sure that applies to large ones as well?
19:49:54 GMT <prsn> redis-benchmark won't tell me, unfortunately. i don't have the ability to set up a whole cluster just to test this, we just have to make a decision about what design to use in elasticache
19:50:29 GMT <minus> well, running redis-benchmark against a local redis instance should give you a ballpark number
19:51:13 GMT <minus> i don't know about hashes and large keys. the size under which it works better is configurable though
19:52:10 GMT <prsn> the main question is whether putting the keys in a hash makes redis have a harder time taking advantage of clustering
19:52:41 GMT <prsn> so benchmarking locally is not really interesting. i guess it might reveal a difference, but that would be pretty surprising.
19:56:48 GMT <prsn> i guess the simplest form of the question is, will redis assign a hash value to a single node, whereas it would distribute top level keys evenly across nodes
19:57:40 GMT <minus> hashes don't make a difference on clustering
19:57:55 GMT <minus> and vice versa
19:58:12 GMT <minus> one hash with however many entries it has is just one key to redis and its clustering
19:58:51 GMT <prsn> ok.  we should definitely just use a lot of top-level keys then, i guess.
19:59:48 GMT <minus> well, if it makes sense to group keys into groups hashes do make sense
21:51:42 GMT <boshia> I am writing an inspec check to test the state of my primaries and secondaries, is there a negative ramification with me passing redis-cli -c SET test_inspec "HELLO"   to check for an OK response on both primaries and the secondary nodes?
21:52:09 GMT <boshia> I know you are supposed to write to primary and read from secondary, but for the purposes of validating everything is operational is this something that would cause a problem?