01:33:54 GMT <felixschl> hi
01:42:52 GMT <felixschl> i've read the in-official as well as the official FAQ first, no luck. is there a mechanism in redis that lets me clean up state redis-side when a client disconnects? say i want a list of online users, so my socket server creates a connection to redis and shares it's state over redis (all clients that connected to it). But should the socket server crash, for instance, is there a way to remove all this state
01:42:54 GMT <felixschl> from redis as well?
02:05:29 GMT <moo-_-> felixschl: hmm
02:05:39 GMT <moo-_-> felixschl: so you would want to execute some script when a client disconnects?
02:05:46 GMT <felixschl> moo-_-: yeah
02:06:07 GMT <moo-_-> felixschl: I feel better place for this logic would be application code, not redis itself
02:06:25 GMT <moo-_-> felixschl: your sockets should have a callback on timeout?
02:06:28 GMT <moo-_-> on server side
02:06:28 GMT <felixschl> moo-_-: but what if the server hosting the application gets it's cable pulled?
02:07:10 GMT <moo-_-> felixschl: if your application loses your connection to database then its usually fatal error and worrying small things really doesn't matter
02:07:12 GMT <felixschl> moo-_-: my usecase: i want to have a reliable way to query all connected sockets to a given room
02:07:34 GMT <moo-_-> felixschl: hmm
02:07:53 GMT <moo-_-> felixschl: I would probably add "last incoming message timestamp" map some sort of of
02:08:02 GMT <moo-_-> felixschl: because with TCP/IP you never know when somebody disconnects
02:08:02 GMT <felixschl> moo-_-: these instances sit behind a load balancer, so there might be a dozen.
02:08:17 GMT <moo-_-> and if timestamp is too old consider the client dead and clean up or ignore
02:08:45 GMT <felixschl> moo-_-: some artifical heartbeat?
02:09:05 GMT <moo-_-> felixschl: heartbeat was the word I was looking for :)
02:09:29 GMT <felixschl> hmm
03:20:17 GMT <felixschl> moo-_-: have you seen this tragic thread by off chance? https://github.com/socketio/socket.io-redis/pull/15
03:35:59 GMT <moo-_-> felixschl: i don't use socket.io
07:44:17 GMT <Ammonium> Hello everyone, I have a question about Redis Sentinnel. What happened if, for any reasons, I have 2 redis master running together after a slave promote ?
07:46:36 GMT <Ammonium> Exemple : I have my redis master on AZ 1a, my slave on AZ 1b. 3 sentinels on each AZ and, a problem, no link between 1a and 1b. Sentinel should promote my slave to master, and I'll have 2 master running together. What happen when the link is retrieve ? One of the master will become a slave ? What about datas ?
08:16:32 GMT <Ammonium> Any ideas ?
08:40:23 GMT <Ammonium> Hello everyone, I have a question about Redis Sentinnel. What happened if, for any reasons, I have 2 redis master running together after a slave promote ?
08:40:26 GMT <Ammonium> Exemple : I have my redis master on AZ 1a, my slave on AZ 1b. 3 sentinels on each AZ and, a problem, no link between 1a and 1b. Sentinel should promote my slave to master, and I'll have 2 master running together. What happen when the link is retrieve ? One of the master will become a slave ? What about datas ?
08:40:28 GMT <Ammonium> Any ideas ?
09:15:17 GMT <minus> one of the masters will become slave again and have all its data overridden with that from the other master
09:16:48 GMT <Ammonium> Hum OK, so I will lose data from one serveur.
09:18:51 GMT <Ammonium> Redis(M) on 1a ------------------ Redis(S) on 1b NETSPLIT Redis(M) on 1a --------X---------- Redis(S) on 1b SO Redis(M) on 1a --------X--------- Redis(M) on 1b THEN Redis(M) on 1a ------------------ Redis(M) on 1b WILL BECOME AGAIN Redis(M) on 1a ------------------ Redis(S) on 1b ?
09:19:57 GMT <Ammonium> (I'm sorry if its not clear, I'm on my first internship and I have few difficulties to understand HA and aws multi az)
11:30:14 GMT <adac> What is the best way to have a master that replicate its stuff to a client periodically?
11:31:19 GMT <adac> *slave not client
11:31:57 GMT <minus> i.e. not continously like with slaveof?
14:31:22 GMT <adac> minus, actually it should be continuously
14:31:53 GMT <adac> Not sure what the replication frequency per default is actually
14:32:34 GMT <minus> normal replication streams continuously
14:33:34 GMT <adac> minus, but it does not slow down the writing process actually, or does it?
14:34:34 GMT <minus> well it is extra load
14:37:39 GMT <rach24py> need a bit help, sudo make install gave me http://pastebin.com/JT6HMDe8
14:39:06 GMT <minus> fun fact: i've never run sudo make install and i consider it very bad practice
14:39:12 GMT <minus> do you not have a package manager?
14:39:18 GMT <rach24py> despite "make test" gave me "\o/ All tests passed without errors!"
14:39:53 GMT <rach24py> yes
14:39:55 GMT <minus> you could run it elsewhere and compare output, or try to make make produce verbose output
14:47:30 GMT <adac> minus, I see yes. but is this write being done synchronously or rather asynchronously?  As I understand it now replication means if I write to the master then the write process is only finished when it is written to the master and the slave, right?
14:48:04 GMT <minus> adac: replication is asynchronous
14:48:30 GMT <minus> i think there is an option to require a change to be replicated to at least x slaves first though
14:48:50 GMT <adac> minus, ok I see thanks. I will look it up better
15:14:23 GMT <RangerMauve> Hullo. If I have replication set up, would writes to slaves get replicated to the master? Or should I somehow ensure that I only send reads to slaves and writes to the master?
15:22:10 GMT <minus> RangerMauve: if you test it out you'll notice that writing to a slave will give you an error (unless you explicitly disabled it. and if you did your changes will /not/ be replicated to the master. that's why it's called the slave)
15:22:26 GMT <minus> there's no master-master replication for redis
15:23:18 GMT <RangerMauve> minus: Awesome! Thanks for clarifying. I guess I'll just have the slaves around for failover for now, and eventually make alll reads go to them.
20:20:09 GMT <felixschl> moo-_-: i figured it would be a longshot (re socket.io)
20:29:36 GMT <moo-_-> felixschl: good good
21:57:27 GMT <astropirate> Hello friends
21:58:30 GMT <astropirate> I have HypeLogLog values in multiple redis instances. What is the best way to read the underlying HLL data structure and write it to another redis instance?
21:58:45 GMT <astropirate> eventually, I would like to perform a PFMERGE operation on them
22:07:51 GMT <astropirate> AHH
22:07:55 GMT <astropirate> I can just use GET and SET
22:09:21 GMT <astropirate> magic
22:19:08 GMT <john_dee> Hey. Does Redis store commands that returned ERR in AOF file?