00:08:03 GMT <harej> I'm currently looking at SSDB for storing a large amount of data. SSDB is a lot like Redis but uses an underlying DB instead of storing everything in RAM. When I told my colleagues this, they suggested I should just use Redis and let Redis swap. Is this actually advised? I know about the VM system that's now deprecated, but I haven't heard of people deliberately using Redis in conjunction with OS swap functionality.
00:13:09 GMT <harej> In my case, I don't think performance is as big a consideration as simply being able to store the stuff is.
00:14:21 GMT <harej> And what I find appealing about using something about SSDB is that it's explicitly designed to use the hard disk, instead of just falling back on the operating system and otherwise not being designed with hard disks in mind.
05:38:10 GMT <kobzar> ьщктштп
05:38:13 GMT <kobzar> morning
08:55:22 GMT <kobzar> I must write the same data in two database (0 and 2 for example). So, what method will be the best?  Use two connection red1= redis.StrictRedis(host='127.0.0.1', port=6379, db=0) and red2= redis.StrictRedis(host='127.0.0.1', port=6379, db=2) and later use it like red1.set(data) + red2.set(data) or use internal command like "select 2" before insert value in next time?
09:02:41 GMT <sev1> Hmm, is there any way to get `RPOPLPUSH` like functionality for Sorted Set? I'm thinking ZADD/ZREM within a Multi/Exec or maybe even within LUA.
10:05:38 GMT <minus> kobzar:doesn't matter if you're not performance-bound. but that sounds like a bad idea anyway
10:06:04 GMT <kobzar> maybe
10:06:05 GMT <minus> sev1: not possible with MULTI, but a Lua script will work
10:17:10 GMT <colo-work> we have an existing redis setup where a central master instance accepts writes, while each application server has a local redis instance that's a replication slave to that central master instance.
10:18:07 GMT <colo-work> the master is a SPOF (the slaves allow for stale reads though), which I'd like to mitigate. I'm eyeing redis-sentinel for that purpose, but I'm not sure if it's OK to have a sentinel'd set of (candidate) masters to also replicate to a set of redis instances that cannot become a master instance
10:18:39 GMT <colo-work> can you enlighten me if that is possible? (or tell me that I'm doing it wrong, if that is the case :))
10:20:51 GMT <minus> colo-work: if they're all on the same LAN it should work; the local/slave redis instances would be managed by the sentinels too, but you can prevent their election to master by setting their slave-priority to 0
10:23:08 GMT <colo-work> minus, all instances are on addresses that share a broadcast domain, if that's what is required
10:23:18 GMT <minus> yeah
10:24:03 GMT <minus> the sentinels should probably go on the application servers, too
10:26:29 GMT <colo-work> my plan is to pair redis instances with sentinel instances 1:1 for each host
10:27:22 GMT <minus> yeah, that's fine; it should be an uneven number however
10:27:52 GMT <minus> not sure if that's really required though
10:28:09 GMT <colo-work> I know. with our current setup, network partitions are both very unlikely and nigh-catastrophic anyway ;)
10:28:50 GMT <minus> heh
10:41:31 GMT <sev1> minus: right, so a lua script of ZADD and ZREM would be an exact replica of RPOPLPUSH?
10:41:45 GMT <sev1> I thought eval would block, is that correct?
10:42:27 GMT <minus> eval is atomic and blocks, yes
10:42:49 GMT <sev1> hmm, so it's not exactly like RPOPLPUSH then or am I missing something?
10:44:52 GMT <sev1> I mean, how does Redis work when it gets `RPOPLPUSH` and a second commend from a different client. Will the first one block?
10:45:27 GMT <minus> RPOPLPUSH isn't blocking
10:45:37 GMT <minus> it won't do anything if the list is empty
10:46:13 GMT <sev1> yeah that's what I am thinking. Remaking it with ZADD and ZREM with eval will block
11:06:28 GMT <adac> Wishlist:  - Docker redis cluster
11:06:30 GMT <adac> :)
11:16:38 GMT <minus> wishlist: no docker please
11:57:01 GMT <sev1> Does every single redis command lock or can you run multiple operations in parallel? The reason I ask is because I am considering between RPOPLPUSH and using lua script to replicate that for a sorted set.
11:57:36 GMT <colo-work> Wishlist:  - no Docker
12:05:53 GMT <sev1> Is a Lua script "stop the world" blocking? Is RPOPLPUSH implemented the same way?
12:06:31 GMT <sev1> I'm guessing since Redis is single threaded, every single operation is "stop the world"?
12:07:12 GMT <xdexter> Hello, My key are expiring before the TTL, it's normal?
12:10:03 GMT <kobzar> No
12:10:40 GMT <xdexter> why this occur? there isn't enough memory?
12:11:20 GMT <xdexter> kobzar, https://pastebin.com/VwWfMkpF
12:38:01 GMT <sev666> Is Redis single threaded in a way that every command run means no other command can possibly read or change anything at the same time. Do I understand this correctly?
12:40:38 GMT <minus> yes
12:41:30 GMT <minus> single threaded synchronous. the only thing that does not run in the same thread is saving to disk; that forks though instead of using a thread
12:46:11 GMT <xdexter> kobzar, can you help me please?
12:47:33 GMT <kobzar> "Hello, My key are expiring before the TTL, it's normal?" how are you check it?
12:48:12 GMT <xdexter> i create a simple PHP to get a session, i access them and get a ID
12:48:29 GMT <xdexter> with ID i get info in REDIS about the key
12:48:54 GMT <xdexter> if i don't refresh the PHP page in ~5 minutes the KEY expire in Redis and my session are lost
12:50:20 GMT <sev666> minus: Alright, so a LUA script itself isn't "heavier" than any other Redis command on e.g list. Just to be clear?
12:50:35 GMT <sev666> Except the content of it might be heavy, but that's not related.
12:51:22 GMT <xdexter> do you understand kobzar? If you need some config file or other info i can send to you
12:51:30 GMT <minus> if you're doing a lot of things in it you will take redis' time
12:51:35 GMT <minus> benchmark it, sev666
12:51:57 GMT <sev666> minus: I am first trying to understand. I think your initial sentence was good. Every command is run synchronous, so it doesn't really matter.
12:52:11 GMT <sev666> Is this true today? The docs I was reading was old.
12:53:20 GMT <kobzar> I doesn't need your config! You say that you key is missed before the expired time is finish. Yes?
12:53:58 GMT <xdexter> kobzar, yes!
12:55:09 GMT <kobzar> Ok! First  - forgot for PHP. Connect to redis with redis-cli (or another client) and make new key with value and expired time by hand.
12:56:09 GMT <kobzar> something like  set new_key some_value     next add expired time   expired mew_key 120   and look the TTL   ttl new_key !
12:56:52 GMT <kobzar> we need find issue place first. This is redis problem or your PHP code.
12:57:24 GMT <xdexter> Ok
12:57:38 GMT <xdexter> set test "Test Session" EX 1440
12:58:11 GMT <kobzar> ttl test ?
12:58:22 GMT <xdexter> EX 1440 set ttl to 1440
12:58:32 GMT <xdexter> ah ok
12:58:33 GMT <xdexter> 127.0.0.1:6379> ttl "test"
12:58:33 GMT <xdexter> (integer) 1368
12:58:56 GMT <xdexter> 127.0.0.1:6379> OBJECT IDLETIME "test"
12:58:56 GMT <xdexter> (integer) 69
12:58:56 GMT <xdexter> 127.0.0.1:6379> ttl "test"
12:58:56 GMT <xdexter> (integer) 1338
12:59:00 GMT <xdexter> sorry
12:59:01 GMT <xdexter> https://pastebin.com/nzEMr75H
12:59:03 GMT <kobzar> so. as we can see all good.  try  get the data
12:59:27 GMT <xdexter> 127.0.0.1:6379> get test
12:59:27 GMT <xdexter> "Test Session"
12:59:39 GMT <xdexter> Ok, we need wait a some minutes
12:59:47 GMT <kobzar> no problem
13:00:17 GMT <kobzar> paste all commands - i do the same in my test redis for compare result
13:01:02 GMT <kobzar> and check the ttl value from time to time     by "ttl test'
13:01:47 GMT <xdexter> Ok
13:01:55 GMT <xdexter> i started a new session in PHP too
13:02:00 GMT <xdexter> to compare expire in both cases
13:06:02 GMT <kobzar> i think problem in PHP
13:06:45 GMT <xdexter> until now none session expired
13:08:20 GMT <kobzar> maybe you use in php another method? for example you set expire value in ms (not in seconds)? As you understand 3600 ms = 3.6 seconds (not 60 seconds)
13:09:09 GMT <xdexter> my php.ini are: session.gc_maxlifetime = 1440
13:10:11 GMT <kobzar> Just look what are you doing. Add new key from you PHP code and check ttl for it in console.
13:10:34 GMT <xdexter> 127.0.0.1:6379> ttl "PHPREDIS_SESSION:vr1k39js3d524rg43e856n8mf7"
13:10:34 GMT <xdexter> (integer) 834
13:10:39 GMT <xdexter> i did this
13:11:26 GMT <xdexter> and i'm waiting the session will be lost
13:15:18 GMT <kobzar> ini_set('session.gc_maxlifetime', 604800);
13:15:18 GMT <kobzar> session_start();
13:15:23 GMT <kobzar> Do you have something like this in your PHP script ?
13:16:12 GMT <xdexter> no, my script is very simple
13:16:13 GMT <xdexter> https://pastebin.com/nA6Z8PvX
13:18:13 GMT <xdexter> it's very strange, the session isn't lost now ;/
13:21:38 GMT <kobzar> like you trouble https://github.com/phpredis/phpredis/issues/586   Just add this parameter in your PHP like this https://github.com/phpredis/phpredis/issues/224
13:24:22 GMT <kobzar> Quick manual https://istvan.xyz/php-sessions-in-redis.html
13:25:39 GMT <xdexter> kobzar, i'm trying simulate the problem now
13:39:41 GMT <mohit> hi I have use case need help how i can i solve with redis
13:45:44 GMT <sev666> minus: Thanks for the help!
17:14:51 GMT <MacWinner> Hi, i have this warning in the logs (which we have fixed), "The TCP backlog setting of 511 cannot be enforced because /proc/sys/net/core/somaxconn is set to the lower value of 128."
17:15:36 GMT <MacWinner> i was trying to understand what the impact of this might be.. we have a server where we get a very occasional "AbortError" from a nodejs redis library.. it seems to be happening more offten as we get more load
17:16:16 GMT <MacWinner> like 1 out of 100000 type of requests.  could the low somaxconn setting be a reason for the "abortError" message?
17:16:44 GMT <MacWinner> we increased to 1024 today and will monitor for the next week.. if we don't see the error again in that time, I'm pretty sure that was it..
17:17:51 GMT <minus> only if you're spawning a /lot/ of new connections to redis
17:18:44 GMT <MacWinner> minus, the weird thing is that my connected_clients=237
17:18:52 GMT <MacWinner> which is past the 128 limit in somaxconn
17:19:12 GMT <minus> that's not what the backlog is for
17:19:12 GMT <MacWinner> so i guess I'm confused by that as well.. is there some relation between connected_clients and the backlog?
17:19:56 GMT <MacWinner> is there any relation between the 2?
17:20:06 GMT <minus> the backlog setting controls how many new connections may be outstanding, i.e how many connections may be waiting for being accept()ed
17:20:13 GMT <minus> no relation
17:20:32 GMT <minus> the total amount of connections is limited by the file descriptor limit
17:26:02 GMT <MacWinner> ahh got it.. what example scenerio would cause the backlog limit to be hit?
17:26:35 GMT <MacWinner> i was trying to simulate that to see if I get teh same AbortError in my nodejs log
17:26:46 GMT <minus> lots of new connections. high CPU usage probably makes it easier to hit that
17:27:25 GMT <MacWinner> ahh.. cool.. very helpful, thanks!
17:39:40 GMT <MacWinner> minus, is there a way to monitor the backlog size at any point in time?  I would love to feed it into my monitoring system to keep an eye on
17:39:51 GMT <MacWinner> like the current queue size i mean
17:40:26 GMT <minus> with some system tools probably
17:40:34 GMT <minus> `ss` might have some info on it
17:41:28 GMT <MacWinner> great, i'll check that out
17:44:09 GMT <MacWinner> had no idea netstat was deprecated in favor of ss..
17:56:00 GMT <MacWinner> minus,ss -lt Send-Q seems to be it.  itneresting read on it here:  inhttps://stackoverflow.com/questions/36466744/use-of-recv-q-and-send-q
17:59:12 GMT <minus> now you know ;)
19:43:24 GMT <ebalsumgo> so… i am hyper confused, i am trying: ./redis-trib.rb check localhost:6379 and getting data back in the form of: M: f857a8600146b53e0a5154a6410a17b8acccdc70 localhost:6379
19:43:24 GMT <ebalsumgo>    slots: (0 slots) master
19:43:50 GMT <ebalsumgo> but when i try and do redis-trib.rb create 1 —replicas localhost:6379 i get [ERR] Sorry, can't connect to node localhost:6379
19:43:57 GMT <ebalsumgo> protected-mode is set to no