00:20:14 GMT anyone around? 00:34:33 GMT Just me and THP. 00:41:58 GMT THP ? 00:42:16 GMT i just had a question that i need help with and i've asked it 24 hours ago and no one has talked since 00:46:55 GMT shisatum, do you know how the LRU method works? 00:50:02 GMT vim /etc/fail2ban/jail.local 00:50:09 GMT i 00:51:23 GMT sorry about that :\ one sec. 00:51:24 GMT wot 00:51:25 GMT lol 00:51:27 GMT lol 00:51:32 GMT if ya dont know it, its fine 00:51:42 GMT no need to avoid me lol 00:51:53 GMT I was working on some server stuff 00:52:10 GMT What was your question? 00:52:11 GMT basically my question was, 00:53:03 GMT if i have some kind of LRU option set for say 1mb. and i have filled up say 998kb and i put in 3kb. and the least recently used object happens to be say 2kb. how does redis deal with the situation? 00:53:11 GMT i know this is oversimplified 00:53:47 GMT if the new object that i put in is bigger than the least recently used object, but the difference being smaller than the second least used, how does redis deal with that? 00:53:51 GMT does it remove the last 2? 00:53:59 GMT if it doesnt, is there a way that i could make it do so? 00:55:32 GMT Unfortunately I wouldn't be able to answer that one, sorry. I'm just in here looking for mentions of issues with Transparent Huge Pages being enabled in RedHat kernels and Redis not liking that. 00:56:19 GMT oh ok damn 00:56:32 GMT your question is much more interesting xD 00:56:47 GMT xD Especially since it causes ecommerce sites to go down randomly. 00:57:01 GMT ohh i see 00:57:13 GMT wait is there a growing number of ecommerce sites using redhat and redis? 00:57:24 GMT The solution is obviously to disable THP, but I've seen a disturbing # of servers lately with it enabled and redis running, to the point where I'm aboutu to write some automation to deal with it. 00:57:45 GMT I'll quit ranting though and let people see your question. :) 00:58:01 GMT Ah, just saw yours. 00:58:31 GMT Well, it seems to happen on Amazon (AWS), too. A lot of AWS customers came from RedHat (RHEL/CentOS) Linux envoronments, so they use similar environments there. 00:58:51 GMT I haven't seen the issue on Debian or Ubuntu. 00:58:58 GMT ah ok 00:59:01 GMT well i use debian 00:59:34 GMT i wouldn't be able to tell you why it fits my use case. it just does. but i doubt there would be difference if i had used another distro 01:00:18 GMT Debian/Ubuntu use THP, too, but supposedly it isn't occurring so much with newer kernels. Either way, you could always test by disabling THP and restarting Redis, and seeing if your CPU usage or latency goes down. 01:00:30 GMT sudo echo never > /sys/kernel/mm/transparent_hugepage/enabled 01:00:58 GMT Developers often use Debian for newer software in the repos, companies often use RedHat for the reliability of their often Older packages. 01:36:22 GMT oh i see shisatum 01:36:22 GMT cool 06:38:15 GMT <_savage> When I run a redis server on localhost, how does it handle multiple users? 17:25:25 GMT hi. a question regarding redis master-slave-replication. what happens after a failover? we have a setup where the slaves (in a cluster) are configured with slave-of 127.0.0.1:6378. stunnel is listening on this port and encrypting the traffic, sending it to the master, where it will be decrypted and redirected to 6370 again 17:25:56 GMT as far as I got it from the doks, this should encrypt the whole communication between master and slave. 17:26:53 GMT but what happens if a slave is populated as master and the master comes back up again? does the slave contact the master, see that he's back and goes back into slave mode or does the master actively establish a connection to the slave to tell him "hey dude, I'm back"? 17:44:47 GMT i'd expect stunnel to fuck over failover. at least that's what the case is with haproxy and sentinel 17:47:07 GMT that's what I assume, too.. 17:48:15 GMT do you have any solution to encrypt the traffic between the nodes? the replication protocol is a binary protocol, but it's a customer requirement that all traffic between the DCs (even though it's a L2 network) is encrypted 17:48:43 GMT we're just using openvpn 17:49:52 GMT client on the slaves and server on the master and a route for the master host IP pushed via dhcp options to the client? 17:52:15 GMT just using it for slave replication without failover here. i'd just go with p2p links between all involved servers/networks with static addresses 17:52:35 GMT ok, I'll keep that in mind 17:52:57 GMT (this is where i'd like to recommend tinc as vpn software because it does just that: a mesh network, but i haven't made good experiences with it) 17:53:28 GMT I guess I have to do some more tests and see if it's possible to encrypt anything with stunnel. redis clustering + master-slave replication with failover 17:53:57 GMT what kind of bad experience with tinc? won't you use it in production? 17:55:09 GMT i've had connection problems with it (connection dropping every minute) but that involved NAT 17:56:55 GMT wow, ok 17:57:16 GMT it's still correct that redis does not support NATted connections, right? 17:57:28 GMT (for replication) 17:57:53 GMT same/similar problem like with stunnel and HAProxy, i guess 17:58:25 GMT don't wanna hear that but you're probably right... 18:09:25 GMT minus: thanks, that was helpful :) 18:15:07 GMT you're welcome 19:31:29 GMT is it correct that redis uses the same port for master-slave replication as it does for the communication within the cluster? (port + 10.000) 19:33:20 GMT no 19:33:48 GMT slaves connect as regular clients on the port that redis is configured to listen on 19:35:50 GMT so a firewall rule to allow tcp on the specified port in both directions would be sufficient for a proper master slave setup? 19:36:07 GMT and the port +10k will only be used if I have a cluster? 19:36:27 GMT for keep-alives and stuff like that.. 19:36:37 GMT yes port+10000 is only a Redis Cluster thing 19:36:59 GMT ok, thanks 20:45:16 GMT hello 20:50:59 GMT is the dump.rdb file where all the data is stored? 20:57:24 GMT seems like it 21:00:06 GMT uyp 21:00:17 GMT yup even