05:27:26 GMT <minus> xrated: how many keys are in your hashes?
09:09:01 GMT <dknight87> hi guys, could someone help me or point me to a link which explains how to setup a 2 node redis db cluster?
09:09:23 GMT <dknight87> I haven't been able to find good documentation
09:32:35 GMT <karthikeayan> Does Redis Sentinel handle fault tolerance?
10:29:20 GMT <badboy_> dknight87: you don't.
10:29:30 GMT <badboy_> dknight87: (well, depending on what you mean by cluster)
10:29:49 GMT <dknight87> badboy_: master, slave
10:29:58 GMT <badboy_> karthikeayan: Redis Sentinel handles automatic failover in case a node dies
10:30:05 GMT <badboy_> dknight87: SLAVEOF master-ip master-port
10:30:18 GMT <dknight87> yup, I've set it up
10:30:23 GMT <dknight87> badboy_: thanks a lot!
10:30:57 GMT <badboy_> dknight87: https://redis.io/topics/replication
10:31:21 GMT <badboy_> that's one of the simplest mode Redis can run in. Keep in mind there's absolutely no handling if any node dies
10:31:28 GMT <karthikeayan> does it mean, it doesn't lose state of the redis node?
10:32:30 GMT <badboy_> karthikeayan: Sentinel is just the monitoring. Data must be replicated from master nodes to slave nodes, so that those slaves can become a new master if their master node dies
10:34:28 GMT <karthikeayan> ok
11:54:14 GMT <mikecmpbll> i'm setting up sentinels for the first time and this sentinel is reporting S_DOWN
11:54:30 GMT <mikecmpbll> although I don't know why. i'd like to test the ping mechanism that it uses to determine if master is down, how can I do that?
11:56:29 GMT <aah> dealing with disk space issues.  can I get rid of temp-* files, in order to free up space for a bgsave?
12:05:04 GMT <badboy_> mikecmpbll: connect to the node via TCP and send PING
12:05:11 GMT <badboy_> aah: yes, those are temporary files
12:05:27 GMT <badboy_> normally Redis should rename to the configured filename if all goes correctly
12:05:50 GMT <mikecmpbll> @badboy_ : thanks! i'll bear that in mind. literally just sorted the issue anyways, i hadn't provided the password of the master correctly to the sentinel.
12:05:57 GMT <badboy_> ;)
12:06:53 GMT <mikecmpbll> this section should mention the `sentinel auth-pass <master> <pass>` config option :p
12:06:55 GMT <mikecmpbll> https://redis.io/topics/sentinel#sentinel-and-redis-authentication
12:07:21 GMT <badboy_> add it and I merge it ;)
12:07:29 GMT <mikecmpbll> 👍🏼
12:16:24 GMT <xrated> minus: sorry just saw message this morning, I did a debug object on one of the hashes and this is the result
12:16:26 GMT <xrated> refcount:1 encoding:hashtable serializedlength:433741 lru:14565491 lru_seconds_idle:1351
12:16:43 GMT <xrated> ive no experience with hashes so idk if that value is large or not
12:28:22 GMT <mikecmpbll> i'm curious how the sentinel config rewrite works. seems to rewrite random portions of my config haha
12:28:51 GMT <mikecmpbll> i had things like `sentinel failover-timeout mymaster 180000` which appear to have been removed
12:29:23 GMT <minus> xrated: HLEN is what i was asking
12:29:38 GMT <badboy_> mikecmpbll: it will remove defaults
12:30:17 GMT <xrated> minus: 4 on avg it looks like
12:30:32 GMT <mikecmpbll> @badboy_ : ahhh.
12:31:26 GMT <minus> xrated: give the redis performance tooling a try then
12:31:34 GMT <minus> latency doctor and what not
12:31:36 GMT <minus> slow log
12:31:40 GMT <xrated> will do
12:31:42 GMT <xrated> ty
12:31:53 GMT <minus> (never used those myself)
12:40:48 GMT <mikecmpbll> the config rewrite stuff makes it kinda tricky to manage redis with configuration automation like saltstack 😓
12:43:52 GMT <minus> iirc cluster fixed that by writing runtime stuff to separate file
12:47:56 GMT <mikecmpbll> okay, i see what's going on. i have sentinels on different vLANs, each connects to the master on a different local IP address. the sentinels keep trying to update eachother with the other master address.
12:48:14 GMT <mikecmpbll> is there a way around this?
12:48:41 GMT <minus> put them on the same VLAN?
12:49:28 GMT <mikecmpbll> :/ i don't want to do that, the whole point i'm setting this up is for resiliency across datacentres
12:49:40 GMT <mikecmpbll> vlan on each datacentre.
12:49:59 GMT <mikecmpbll> ah maybe there's some networking fu i can do. i suck at that stuff.
12:50:05 GMT <minus> you set up sentinels cross-datacenter?
12:50:14 GMT <mikecmpbll> minus : aye.
12:50:22 GMT <minus> sounds questionable
12:50:25 GMT <mikecmpbll> why?
12:50:49 GMT <minus> just does
12:50:57 GMT <minus> i guess it depends on what you use it for
12:51:26 GMT <minus> a netsplit with that is gonna be annoying as fuck
12:51:38 GMT <minus> assuming both sides write
12:53:27 GMT <mikecmpbll> all our infrastructure is replicated across these two datacentres. load balancers, databases, everything
12:57:04 GMT <badboy_> no, Redis doesn't really deal easily with different IPs for the same instance
12:57:25 GMT <mikecmpbll> @badboy_ : thanks
13:00:49 GMT <mikecmpbll> eh i'm an idiot anyway i just need to add the sentinel hosts to the redis vlan.
13:04:10 GMT <minus> mikecmpbll: so, how does this work, does only stuff in one DC do actual work and the other is on standby or do both work and only work with the redis master and thus are blocked from working while a netsplit is occurring?
13:05:12 GMT <mikecmpbll> not sure if i understand the question necessarily, but i'm just going for a master/slave scenario with failover
13:05:27 GMT <mikecmpbll> so yeah, one server will be doing all the work at any one point.
13:05:46 GMT <mikecmpbll> if it goes down, it switches to the other one at the other DC.
13:05:55 GMT <minus> ah okay then
13:06:15 GMT <minus> so pure failover
13:06:29 GMT <mikecmpbll> aye, no sharding or load balancing or anything like that
13:06:47 GMT <mikecmpbll> we could, but we don't need that.
14:46:17 GMT <JohnnyRun> hi all. Is it possible to disable persistence at all on redis? I'm using 2.8.19 on centos. here is my conf dump: https://gist.github.com/johnnyrun/af63ddb91385abdff269c7c4b023b0fd
14:46:47 GMT <JohnnyRun> I have 'save ""' in the conf but I find dump.rdb on filesystem.
14:55:29 GMT <badboy_> do you have slaves attached?
14:55:44 GMT <badboy_> oh it is a slave
14:57:31 GMT <badboy_> Then no, on initial sync a dump will be saved
14:58:13 GMT <JohnnyRun> ok badboy. so it is slave-related?
14:58:29 GMT <badboy_> yes
14:58:53 GMT <JohnnyRun> Do you know if it is the same on newer versions?
14:59:43 GMT <badboy_> yes, it is on all versions
14:59:48 GMT <badboy_> there's a diskless mode for the master
14:59:57 GMT <badboy_> but the diskless-mode for slaves is not merged
15:00:17 GMT <JohnnyRun> ok. Thanks a lot badboy_
15:22:19 GMT <minus> .qrand album:largo
15:22:24 GMT <minus> whoops
16:10:31 GMT <tessdev> Hi,We are facing redis connection timeout even after upgrading to aws m4.xlarge
16:12:37 GMT <minus> you have to lower the frequency of saves