09:33:11 GMT <catphish> i'm looking for a solution for a multi-tenanted key value store, i understand redis isn't great for multi-tenanted use, but are there any ways to make it so? i wanted to avoid having to run lots of instances, perhaps a proxy?
09:34:14 GMT <minus> catphish: run multiple instances. if you use a proxy you still have no isolation unless you prohibit the usage of scripts and replication and maybe other things
09:35:06 GMT <catphish> minus: i was thinking that a proxy would specify a database and prevent the use of most commands (only allowing those that relate to data storage in the current database)
09:35:31 GMT <catphish> running multiple instances is a headache, but starting to seem like the direction i need to go in
09:38:26 GMT <minus> it's not really more of a headache than setting up a proxy
09:40:08 GMT <minus> even if you have a high rate of tenant changes
09:40:23 GMT <catphish> it's just more moving parts having to assign port numbers and ensure instances are started and stopped reliably when needed
09:40:44 GMT <catphish> its the high rate of change that worries me
09:41:02 GMT <minus> well, you have to automate it one way or another
09:41:21 GMT <minus> though changing a proxy configuration is arguably easier
09:43:04 GMT <catphish> i would certainly think regenerating a proxy config was easier, even better would be writing a proxy that uses a database to route connections, but i don't know the protocol well enough to write something really
09:48:38 GMT <minus> the protocol is simple
09:48:55 GMT <minus> but writing a good proxy probably isn't so easy
09:49:39 GMT <minus> the basic stuff should be simple, but edge cases make it hard to get completely right
09:49:46 GMT <catphish> makes sense
09:50:28 GMT <minus> there's already some proxies out there, dunno if one of them works for your case
09:50:59 GMT <catphish> i'll have a look
17:06:16 GMT <drpsycho> We're testing redis 4 with sentinel and we constantly have partial sync issues
17:06:28 GMT <drpsycho> When a master dies, a new master is elected and the failover happens smoothly, but when the old master comes back online as a slave, it is unable to do a partial sync
17:07:17 GMT <drpsycho> The error is either that the new master doesn't recognize the old replication ID or the offset requested is out of range
17:07:40 GMT <drpsycho> I've tried to increase repl-backlog-size to 1000m but it doesn't seem to help
17:08:06 GMT <drpsycho> And this is with basically no load on the redis instances. The only thing writing to redis is sentinel itself
17:08:22 GMT <drpsycho> Any tips for how I can overcome this?
17:10:10 GMT <drpsycho> I've also tried with this patch: https://github.com/antirez/redis/pull/3867
17:10:14 GMT <drpsycho> It doesn't seem to help
21:28:39 GMT <redis_noob> hi guys, I'm running 3 redis nodes behind a VIP and when I get redirected, redis redirects using the IP address instead of the VIP
21:29:02 GMT <redis_noob> the ip addresses aren't accessible externally. is there any way to get redis to use the VIP hostname instead?