10:36:08 GMT <netcrime> Hey guys I get the error: PHP Warning:  PHP Startup: Unable to load dynamic library '/usr/lib/php/20151012/redis.so' - /usr/lib/php/20151012/redis.so: cannot open shared object file: No such file or directory in Unknown on line 0
10:36:16 GMT <netcrime> How could I solve it? Using ubuntu 16.04 PHP 7
10:57:49 GMT <minus> i'd say try reinstalling the php redis extension
11:38:18 GMT <p1rate> pecl install redis ?
11:38:32 GMT <p1rate> or apt-get install php70-redis (or the php version you're using)
18:06:30 GMT <dh____> Can anyone help me figure out how to have a Redis master not talk to Sentinel? We are running 1 master + 3 slaves in a Kubernetes cluster/namespace with 3x Sentinel nodes.
18:06:55 GMT <dh____> We are trying to deploy a single master outside of the redis cluster, in the same namespace. It keeps talking to sentinel and becoming a slave instead of independent master
18:07:15 GMT <minus> does that do any kind of NAT?
18:07:50 GMT <dh____> for internet connectivity, between nodes is overlay L3 network
18:07:59 GMT <dh____> which has direct connectivity
18:08:05 GMT <zetab> dh____: i have tried months to figure out how to deploy redis cluster to kubernetes
18:08:13 GMT <zetab> the problem is that there is not really easy way to do that
18:08:21 GMT <dh____> We have the cluster running fine
18:08:23 GMT <zetab> imo redis could do something to that problem
18:08:29 GMT <zetab> dh____: yes but what about failovers etc?
18:08:35 GMT <dh____> It's just trying to add a single node in addition to the cluster, it joins the cluster automatically
18:08:39 GMT <zetab> if you shutdown master what will happen
18:08:39 GMT <dh____> instead of staying independent
18:08:55 GMT <dh____> zetab, they have new master election
18:09:06 GMT <zetab> and how services knows what is the new master?
18:09:10 GMT <zetab> you need modify service labels at least
18:09:15 GMT <dh____> hehe
18:09:17 GMT <dh____> that was the fun part
18:09:30 GMT <zetab> like you can see redis is not really very cloud ready :D'
18:09:45 GMT <dh____> We have a liveness check that checks the actual health to decide to kill the pod
18:10:07 GMT <zetab> i have decided that i will just run single redis instance in kubernetes with PV attached
18:10:13 GMT <dh____> but then we have a readinessprobe that tests whether it is the master, by querying for "Role: MASTER"
18:10:21 GMT <zetab> we have downtime if the pod dies, but that is life
18:10:23 GMT <dh____> and the slaves stay in an unready (but alive) status
18:10:26 GMT <dh____> until elected master
18:10:51 GMT <zetab> ok
18:11:22 GMT <dh____> we had to reduce the master election time on the sentinel config
18:11:23 GMT <zetab> it would be really nice to see good solution for that
18:11:32 GMT <zetab> statefulset redis cluster
18:12:22 GMT <dh____> our readinessProbe looks like this, FYI: sh -c "/usr/local/bin/redis-cli -h $(hostname) INFO | grep 'role:master'"
18:14:28 GMT <zetab> k
18:15:34 GMT <zetab> dh____: hmm so basically only one pod is serving request at the time?
18:15:42 GMT <dh____> yeah
18:15:52 GMT <dh____> its very limited for our setup
18:16:09 GMT <dh____> since the pods would not be ready, we cannot tie them to a second read-only service
18:16:17 GMT <zetab> i am just thinking is it worth of it, to make rocket science instead of running single redis with PV :D
18:17:07 GMT <zetab> redisconf 2017 is end of this month. Hopefully someone will do good presentation about redis in kubernetes
18:17:40 GMT <minus> #JustOpsThings
18:31:38 GMT <zetab> DevOpsThings* :)
18:32:19 GMT <minus> but that's just ops stuff
18:33:11 GMT <minus> do you have a team providing the kubernetes infrastructure for you?
18:33:27 GMT <minus> or do you also maintain that
18:34:38 GMT <zetab> we maintain that
18:38:43 GMT <minus> and you have time to dev stuff?
18:45:38 GMT <zetab> not much :D
18:46:13 GMT <minus> then i wouldn't quality it as devops
18:46:15 GMT <minus> ¯\_(ツ)_/¯
18:53:17 GMT <nemothekid> Hi, I'm on redis 2.8.23 - and I'm trying to debug a performance issue. I use hashes a lot and I'm trying to understand why sometimes, a HSET command will take ~100ms in the slow log.
23:30:13 GMT <wu_> hi