01:11:28 GMT <raspado> when installing redis, should one go with redis 2.8 what is in yum/epel or compile it from source (3.2)
05:46:33 GMT <minus> Freman: keep in mind that redis has buffers for client connections. dunno if pubsub messages are copied for each client before sending though
05:47:08 GMT <Freman> actually trying to track down intermittent errors.
05:47:39 GMT <Freman> we like to blame phpredis but I've replicated it in golang now :(
05:49:53 GMT <Freman> admittedly, our redises are thoroughly abused but can you think of any reason this would happen (it's a tight loop, time is in seconds on the left)
05:49:54 GMT <Freman> http://pastebin.com/8WSgkqbz
05:50:33 GMT <Freman> all it's doing is connecting (issueing a ping to force a lazy_connect to connect) and then issuing a second ping to track rtt
05:50:38 GMT <Freman> and then disconnecting
05:53:02 GMT <minus> looks terribly unstable
05:54:03 GMT <minus> the connection between the machines is fine?
05:55:14 GMT <minus> am i reading this right, a timeout happens every 5 seconds?
05:55:35 GMT <minus> and the iterations between each timeout vary greatly?
05:57:53 GMT <minus> ah, it's 5s timeouts
05:59:19 GMT <Freman> yeh the connection is fine, I've got a ping and it isn't exactly stable but it hasn't exceeded 23 ms (100ms interval) - rtt min/avg/max/mdev = 0.147/0.461/22.290/1.634 ms
05:59:56 GMT <Freman> I've got a tcp dump that shows the syn - syn/ack time being 1.0028 seconds on one connection in every... 6-800 or so
06:00:43 GMT <minus> any of the machines at its limit? or maybe some network hardware inbetween?
06:01:09 GMT <minus> you could also enable slowlog in redis and have a quick look on that
06:02:32 GMT <Freman> network wise no, both machines have bonded dual gigabit links, I've got another test that makes connections and smashes the living hell out of pubsubs, can handle 1000 simultaneous connections sending 250kb messages
06:02:49 GMT <Freman> the machine is dedicated to pubsub
06:03:09 GMT <Freman> does slowlog show pubsubs?
06:03:13 GMT <minus> dunno
06:03:48 GMT <minus> what's the backlog setting in redis? maybe that's too low. though i can just guess
06:03:58 GMT <minus> and gotta go to work now, afk
06:04:01 GMT <Freman> (in fact of all the abused redis servers we have, this one is the least abused)
06:06:28 GMT <Freman> tcp-backlog 5000, maxclients 50000, client-output-buffer-limit pubsub 32mb 8mb 60
06:24:33 GMT <Freman> yeh the latency tools give me bumpkiss
06:50:00 GMT <Freman> went and grabbed the latest version of redis, default config (+ listen 0.0.0.0 and protected-mode no) and still a problem on bare metal even ditching docker
06:58:05 GMT <Freman> same test on a ubuntu box and coreos :(
07:18:48 GMT <Freman> so to docker on coreos, coreos, ubuntu all timeouts. to localhost ubuntu no timeouts
07:18:57 GMT <Freman> any one know of any kernel tuning parameters?
07:53:49 GMT <Freman> running over rmux removes the timeout (probably due to the already connected connection pooling)
07:53:58 GMT <Freman> but that doesn't support subscribe
12:35:49 GMT <kbni> Hi all
12:36:46 GMT <kbni> Can someone recommend their redis-web management platform of choice? Fairly simple deployment, but I'm after something I can easily document less technical users setting/updating keys etc.
13:31:17 GMT <jeffreylevesque> how do i make sure redis starts every time i boot my centos 7
13:36:58 GMT <scottc1> jeffreylevesque: systemctl enable redis.service
13:40:06 GMT <jeffreylevesque> scottc1: do i still need to `redis-server` to start the server ?
13:41:04 GMT <scottc1> jeffreylevesque: systemctl start redis.service should do it
13:41:26 GMT <jeffreylevesque> scottc1: do i need to run both, or just the last command?
13:42:24 GMT <scottc1> jeffreylevesque: systemctl enable will configure the default runlevels to start it at boot. systemctl will start it immediately
13:42:37 GMT <jeffreylevesque> thank you!
13:43:15 GMT <scottc1> jeffreylevesque: No problem
14:35:35 GMT <raspado> hi all, quick Q, once I make/make install redis, can I delete the install directory that is extracted via tar
14:36:45 GMT <raspado> once it has already created the bin directory
15:10:21 GMT <vt> is anyone using redis with logstash/elasticsearch?
15:38:00 GMT <raspado> does HA cause any noticeable delays?
16:11:19 GMT <[0xAA]> does Redis keep the whole key-value set in memory
16:11:35 GMT <[0xAA]> Because I don't need all in memory and I fear it'll take up all of my ram
16:15:20 GMT <sknebel> [0xAA]: redis is an in-memory datastore, so yes it does.
16:19:53 GMT <[0xAA]> so it'll take up all my RAM
16:31:03 GMT <borei> hi all
16:33:09 GMT <borei> im new one with redis, trying to build HA cluster using sentinel, i modified config file according my network environment, started redis-server and redis-sentinel and gor sentinel.conf file modified ???
16:33:51 GMT <borei> why does it happen ?
16:47:31 GMT <borei> and redis.conf also modified
16:47:47 GMT <borei> no doubt im missing something important
16:47:52 GMT <gajus> What happens when redis slave is allowed to write?
16:48:01 GMT <gajus> i.e. would that completely cut off sync
16:48:14 GMT <gajus> or would that just stop sync of specific keys
16:48:16 GMT <borei> i don't know yet
16:48:25 GMT <borei> my config modified somehow
16:48:56 GMT <borei> slaveoff settings changed from 10.200.xxx.xxx to 127.0.0.1
16:49:12 GMT <borei> but why ???
16:53:14 GMT <borei> i stop sentinel on the slave server, everything looks good
16:53:55 GMT <gajus> does `slave-read-only no` cause reads against master on each hash lookup?
16:55:38 GMT <borei> i have slave-read-only yes, so it's in read-only mode
16:56:13 GMT <gajus> borei I am not talking to you. I am asking a question.
16:56:25 GMT <gajus> I haven't even seen your question; I have joined the channel just now.
16:56:39 GMT <borei> :-)
16:56:48 GMT <gajus> Back to the question. The reason I am thinking it will cause a hash lookup against the master is this paragraph
16:56:50 GMT <gajus> 'Normally slave nodes will redirect clients to the authoritative master for the hash slot involved in a given command, however clients can use slaves in order to scale reads using the READONLY command.'
16:57:26 GMT <gajus> it sounds like `slave-read-only yes` will cause each lookup against the master... then whats the point of the slave?
16:59:58 GMT <gajus> I have asked here http://stackoverflow.com/questions/39622451/does-setting-slave-read-only-no-will-make-slave-confirm-every-hash-lookup-with
17:18:48 GMT <sknebel> gajus: http://redis.io/commands/readonly - the client has to opt-in to potential stale reads, but if it does the master is not queried
17:22:52 GMT <sknebel> gajus: sorry, ignore me, misread your question.
18:11:53 GMT <minus> borei: sentinel rewrites its own and the slaves' config files
18:12:55 GMT <minus> gajus: disabling slave readonly causes writes to a slave not to fail with READONLY. the changes you make are local to the slave and could be overridden by changes from the master
18:13:54 GMT <borei> minus: i found it, it happens when slave is losing connection to master
18:15:06 GMT <minus> borei: if you want ephermal cache and don't need the master/slave setup for that i suggest you spawn a second (set of) redis instance(s)
18:16:06 GMT <minus> oh, this is cluster stuff
18:16:17 GMT <borei> yeah, i need HA
18:17:07 GMT <minus> if you just need HA but no partitioning, redis-sentinel may be more appropriate, though i'm not sure on that; i haven't used cluster, only sentinel
18:18:02 GMT <borei> im trying to use sentinel
18:18:07 GMT <borei> setup is dirt simple
18:18:17 GMT <minus> then the readonly command does not apply
18:18:28 GMT <borei> 2 redis instances, 2 sentinel
18:18:33 GMT <minus> you should use 3 sentinels
18:18:47 GMT <minus> otherwise they could have problems coming to a decision
18:19:08 GMT <borei> yep i know about it, once in production i'll have 3 of them
18:19:31 GMT <borei> but, i don't understand why slave is losing connection to master
18:20:05 GMT <minus> if you fucked up your sentinel configs once it's easiest to just wipe all redis and sentinel configs and start over
18:20:23 GMT <borei> i did it
18:20:43 GMT <borei> started master and slave instances, got initial sync
18:20:45 GMT <minus> `slave-read-only no` then allows you to write to slaves (has no impact whatsoever on the master)
18:21:12 GMT <borei> 30 seconds down the road master and slave lost connection
18:24:37 GMT <minus> sounds odd
18:24:49 GMT <borei> 11713:S 21 Sep 11:22:45.691 * MASTER <-> SLAVE sync: Finished with success
18:24:58 GMT <borei> 11713:S 21 Sep 11:23:51.308 # Connection with master lost.
18:25:07 GMT <borei> drives me nuts
18:25:08 GMT <minus> we've had problems like that with extra slaves attached to HAProxy
18:25:21 GMT <minus> sounds odd
18:26:11 GMT <borei> can network issue, not sure yet
18:26:17 GMT <borei> can be ^^^
19:10:49 GMT <borei> i've nether seen software that rewrite it's own configuration file
19:11:16 GMT <borei> don't know yet - is it good idea or not
19:11:31 GMT <borei> but it makes certain level of diffuclties
22:09:07 GMT <Freman> https://snag.gy/dOZD2f.jpg <- our client connecting to a redis-server with a 1s timeout :(
22:09:22 GMT <Freman> it looks the same with a 30s timeout, but takes longer
22:25:44 GMT <ayee> > keys cc4d3106-e727-411f-b0c7-2af59530e550\n 1) "cc4d3106-e727-411f-b0c7-2af59530e550"
22:26:01 GMT <ayee> > get cc4d3106-e727-411f-b0c7-2af59530e550\n (error) WRONGTYPE Operation against a key holding the wrong kind of value
22:26:13 GMT <ayee> I'm confused.. I'm never used redis-cli .. how can I get the value of this key?
22:26:25 GMT <ayee> I can see the key exists, I don't understand how to just dump the value of this key
22:26:28 GMT <Seventoes> what type of value is cc4d3106-e727-411f-b0c7-2af59530e550?
22:26:33 GMT <Seventoes> "type cc4d3106-e727-411f-b0c7-2af59530e550"
22:26:43 GMT <ayee> hash
22:26:56 GMT <Seventoes> use the hash-related functions to get the value then, "get" is for simple string values
22:27:15 GMT <Seventoes> http://redis.io/commands#hash
22:28:00 GMT <ayee> sweet, hgetall is what I wanted. Thanks for the help @Seventoes, I appreciate it.
22:28:07 GMT <Seventoes> not cli-specific btw ;) same rules for client libraries
22:35:45 GMT <borei> completely lost direction
22:36:31 GMT <borei> i have 2 redis servers (master slave) and 3 sentinel
22:36:47 GMT <borei> when i do start redis servers everything looks good
22:37:02 GMT <borei> on server is master. second slave
22:37:31 GMT <borei> starting sentinel make things happen
22:37:59 GMT <borei> master server becoming slave, pointin to it's own ip as to master
22:39:12 GMT <Freman> Ok, I've proven our problems aren't redis specific but actually network/kernel (I wrote a tiny redis server :D)
23:07:35 GMT <Janos> hey there guys, we have a website and are think about implementing redis as http session store and general cache, I have been reading on the redis docs and I see there are multiple options for having redis on multiple servers, so far I have found, replication, cluster and sentinel and there are probably more
23:08:27 GMT <Janos> I was wondering if there is any documentation that compares these different configurations and present the pros and cons of each
23:09:40 GMT <Janos> and hopefully some common scenarios for each of those setups
23:28:28 GMT <Freman> connections seem to not always work (Don't have a better explanation) - https://gist.github.com/freman/90dec028f72216837da555ac3594b52a - and I've captured this in wireshark https://snag.gy/dOZD2f.jpg
23:29:20 GMT <Freman> I'm not transferring much data (8 bytes) just connecting, ping, pong, disconnecting
23:29:51 GMT <Freman> any idea what might be causing this? it's a dedicated lan, bonded dual gigabit nics
23:31:29 GMT <Freman> using kernel-4.6.3
23:45:26 GMT <Freman> ach, this was meant to be the linux channel lol, I blame limechat or having one text box across all channels