07:07:20 GMT <freddo> Hi. I'm looking for advice on a 2 server hot standby setup. I'm implementing it by hand because Sentinel and Cluster don't seem to support a 2 server - and I've only got 2 CPUs . Any good example client code/doc for thIs scenario, to handle detection and switchover to standby server? Thx.
07:38:32 GMT <minus> freddo: sentinel works with 2 redises, but needs at least 3 sentinels (so at least one would have to be external)
07:39:06 GMT <minus> you can't really do failover with two servers because if the connection between them fails both think they should be master
07:42:18 GMT <freddo> minus: thanks, that's what I understood, but this is an isolated system and I don't have a 3rd CPU to run sentinel on. I'm familiar with split brain problem, but I have a hardware line between the two servers to reduce probability of that occurring.
07:45:33 GMT <minus> freddo: in principle you just have to CONFIG SET SLAVEOF NO ONE on the salve when the master goes away
07:58:31 GMT <freddo> minus: Thanks. Do you know if ping is a sufficiently reliable way of detecting the failure?
07:59:28 GMT <minus> assuming you're not running out of memory: yeah
08:00:02 GMT <freddo> minus: great - thanks.
08:00:21 GMT <minus> and if the instance is just starting it'll return an error (LOADING Redis is loading the dataset into memory) if you're trying to read something from it
14:08:26 GMT <adac> Hi there!
14:08:47 GMT <adac> I was wondering if redis has some kind of transaction log somewhere?
14:14:14 GMT <badboy_> no
14:14:32 GMT <badboy_> what are you trying to do?
15:11:31 GMT <adac> badboy_, soemthing mysteriously went wrong on our staging server, some stuff got deleted and I thought I could maybe reproduce it
15:12:12 GMT <adac> not reproduce it but of look up some transaction logs to find out what and if somehting happened
15:12:17 GMT <badboy_> if you have AOF enabled, you could^^
15:12:55 GMT <adac> badboy_, how would that work with AOF?
15:13:05 GMT <adac> for now I only have rdb
15:13:13 GMT <adac> but soon this needs to change anyways
15:13:18 GMT <badboy_> the AOF is an appendonly log of all write commands
15:13:22 GMT <badboy_> (sort of a "transaction log"
15:13:34 GMT <adac> badboy_, ok I see so all would be there yes
15:13:51 GMT <badboy_> unless rewritten, yes
15:13:59 GMT <adac> would be handy. we should really use that for this redis instance where we have crucial data
15:14:14 GMT <adac> rdb to AOF can be done with no problems or?
15:15:51 GMT <badboy_> you can actually enable both ;)
15:16:23 GMT <badboy_> see https://redis.io/topics/persistence
15:17:01 GMT <badboy_> CONFIG SET appendonly yes
15:17:02 GMT <adac> badboy_, oh thats nice so I just have to enable it
15:17:12 GMT <adac> thanks man
17:10:56 GMT <crayon>  I'm trying to resolve a redis connection issue. U've determined the cert is being rejected because of name validation. To correct it, I need to regenerate with either: CN changed, SAN entry for the VIP, or wildcard.
17:10:58 GMT <crayon> any suggestions?
17:11:15 GMT <badboy_> crayon: what cert?
17:14:11 GMT <crayon> it's internally generated, signed by the CA we use for the redis db
17:14:33 GMT <crayon> for redis clients
17:14:39 GMT <crayon> (auth)
17:33:21 GMT <minus> Redis doesn't support cert auth nor encryption though
17:53:00 GMT <crayon> redis.clients.jedis.exceptions.JedisConnectionException: The connection to 'project21.zelda.io' failed ssl/tls hostname verification.
17:53:19 GMT <crayon> it's not needed for "public" TLS, but is used to provide trusted TLS (through possession of the CA, which is preshared), and client certs for authorization (again, signed by CA).
18:00:37 GMT <badboy_> well, fix your cert then ;)
18:00:53 GMT <badboy_> or (DONT DO THIS AT HOME) disable hostname verification (DONT DO THIS)
18:10:58 GMT <alex88> hello everyone, I've this lua script and the logs of `monitor` https://gist.github.com/alex88/4ad4f73977332d3341d05f99e71c8a9d
18:11:28 GMT <alex88> as you can see the only operations done on foo are a RPUSH, LPOP and then LINDEX which is the script on line 14
18:11:52 GMT <alex88> however, I see `attempt to concatenate local 'next_message' (a boolean value)`
18:11:58 GMT <alex88> how can it be a boolean?
18:12:19 GMT <alex88> if I first rpush and then lpop, the list should be empty right? so it should return nil
18:12:28 GMT <minus> have you tried redis-cli --ldb?
18:13:08 GMT <alex88> `Options --ldb and --ldb-sync-mode require --eval.` ?
18:14:03 GMT <alex88> oh ok, well, I was calling that function from our unit tests of our app, not directly via cmdline
18:14:11 GMT <alex88> let me move the script and run the debugger
18:15:51 GMT <alex88> yeah next_message = false
18:17:29 GMT <alex88> I don't see why though
18:17:45 GMT <alex88> I pushed only strings
18:19:37 GMT <minus> maybe you found a bug
18:19:53 GMT <badboy_> a nil response is converted to a bool, isn't it?
18:19:58 GMT <badboy_> > Redis Nil bulk reply and Nil multi bulk reply -> Lua false boolean type
18:20:01 GMT <alex88> oh, that could be the reason
18:20:08 GMT <alex88> bulk reply?
18:20:41 GMT <alex88> anyway yes, redis.call("LRANGE", "bar", 0, 1000) shows {} but LINDEX is false
18:21:30 GMT <badboy_> bulk reply is just the Redis Protocol's name for a byte array
18:21:43 GMT <alex88> ok gotcha
18:22:00 GMT <alex88> I was just doing LINDEX on redis-cli and wondering what that was nil and lua was false
18:22:36 GMT <alex88> as I've switched from LRANGE + next(next_message) to LINDEX and got into that, thanks a lot anyone
18:22:40 GMT <minus> why doesn't it return nil for a nil bulk reply and an empty table for nil multibulk :|
18:22:56 GMT <badboy_> leegacy :D
18:26:53 GMT <alex88> :D
18:27:25 GMT <minus> time to make a redis.call2
18:37:58 GMT <alex88> or just move to redis 4 :D
18:38:15 GMT <minus> does redis 4 fix it?
18:38:19 GMT <minus> i doubt it
18:38:23 GMT <minus> because compatibility
18:44:59 GMT <alex88> I don't know, I just said that since a major version could change these things