04:59:54 GMT <hwkng> what is the term I want to google for for "I want to figure out how to setup multiple machines; have them run redis, and split them by key"
08:11:15 GMT <redis092> Hello guys
08:11:38 GMT <redis092> I have a small question about Redis performance, currently I have around 200K keys and it takes around 20s to load them using "KEYS *"
08:11:49 GMT <redis092> does that look correct to you?
09:28:59 GMT <nnog> Hey guys, for those clued up on the latency findings here: http://antirez.com/news/84, I have a Q. Lets say I was forced to have THP enabled. Would it be better to use libc malloc over jemalloc to avoid these latency spikes?
09:30:31 GMT <nnog> As I understand, it will prevent the big spikes due to not using any anonHugePages, but suffer worse fragmentation...
09:31:34 GMT <jamespedwards42_> Is there any way to add a slave node to Redis Cluster but prevent fail-overs from happening? The purpose is scale up read requests and avoid lost writes to a network partitioned master.
09:31:59 GMT <nnog> Perhaps I made the wrong conslusion, though. Main Q is: does THP hurt when using libc malloc?
09:35:47 GMT <nnog> I'm concerned that because I'm using RDB persistence, and writes generally come all at once, that the first write of the large burst of writes will trigger a fork, and then, since I can't disable THP, there will be a big latency spike.
12:21:36 GMT <N3sh108> heyhey! I am planning to deploy my website with a session. Since the website can be scaled horizontally, I thought to use Redis as session store. That’s fine but I was wondering: what is the best way to have it distributed?
12:21:58 GMT <N3sh108> I am building my website around Kubernetes, could the Redis containers also be deployed like that?
12:22:15 GMT <N3sh108> Since they need to have persistent data, I am not sure how I would do it though
14:28:55 GMT <minus> impatient folks
14:31:42 GMT <ahfeel> minus: :<
14:32:37 GMT <minus> jamespedwards42_: slave-priority 0 maybe, dunno if redis cluster cares about that though (read the manual with regard to that)
15:32:43 GMT <jamespedwards42_> minus: Thanks, I will give it a try, I assumed slave-priority was only used by sentinel, not Redis Cluster.
15:41:25 GMT <jamespedwards42_> minus: it looks like it is only used by sentinel: github.com/antirez/redis/blob/unstable/src/server.h#L969... searching the code for slave_priority shows it is only used by sentinel.c
15:42:01 GMT <minus> well that sucks
15:43:42 GMT <jamespedwards42_> minus: :) eh, time to go open another not really that important feature request
17:32:44 GMT <jamespedwards42_> minus: just in case your interested in the Redis Cluster slave-priority=0 issue: https://github.com/antirez/redis/pull/3069
19:19:16 GMT <minus> not sure why that needs an extra flag
19:24:15 GMT <minus> your java client library looks odd, jamespedwards42_. why the requirement for lambdas?
19:25:59 GMT <jamespedwards42_> it manages providing the client for you as well as the retries and exception handling.  so there is some inversion of control taking place. lambdas are great for providing that in a clean way.
19:27:10 GMT <minus> retrying would work without them as well though. unless you have more than one command in the lambda, i guess
19:28:29 GMT <jamespedwards42_> yes, it is much more flexible this way.
19:30:44 GMT <minus> i don't quite see it yet completely, i'll put that on my list of things to look over for monday :)
19:32:49 GMT <minus> any particular reason for doing client.sendCmd(Cmds.GET, …) over client.get(…)? i find latter nicer because it gives you the docs with the function outline already
19:34:31 GMT <jamespedwards42_> but you end up waiting for the project maintainer to add and release new commands
19:35:02 GMT <jamespedwards42_> also, to me it feels bloaty and a pain to maintain
19:35:52 GMT <jamespedwards42_> this way, it encourages the user to create whatever cmd they want (modules, etc), and also deserialize the response however they want.
19:37:31 GMT <minus> you aren't taking care of serialization at all?
19:38:12 GMT <jamespedwards42_> yes, of course, see the readme examples and tests.  but beyond basic deserialization it doesn't look you into returning something as a HashMap for example
19:38:17 GMT <jamespedwards42_> like you see in other clients
19:38:20 GMT <minus> it's super useful if it work, but also super painful if it doesn't. i've had that with Spring's Redis thingie
19:38:38 GMT <jamespedwards42_> *lock you into
23:44:58 GMT <raz> hmm, is the redislabs pricing a joke? :/
23:45:13 GMT <raz> 64GB instance for $5011/mo?!