This version (2017/05/27 13:44) is a draft.
Approvals: 0/1

[08:51:00] <mweb84> hi there, do I need a system property like “vertx.clusterManagerFactory” in vert.x 3x clustering?

[22:55:42] <skullcrasher> is there a number on max verticles in a vert.x instance and also on max nodes in a vertx cluster?

[22:56:13] <AlexLehm> no, i don't think there is

[23:01:39] <temporalfox> skullcrasher sky is the limit

[23:01:51] <skullcrasher> hehe ok :D

[23:01:55] <skullcrasher> thx guys

[23:02:57] <skullcrasher> thought more about how many are feasable until the clusters get unusable (akka e.g. can have about 2500 cluster nodes until it crashes)

[23:03:22] <skullcrasher> but I have no problem if there is no know limitation :)

[23:03:46] <skullcrasher> also pretty unlikely I will reach that limit in near future :D

[23:04:09] <temporalfox> skullcrasher in vert.x it depends likely on the cluster manager implementation

[23:04:14] <temporalfox> ah there are limitations

[23:04:29] <temporalfox> it's just that we haven't tested ourselves the hard limits

[23:04:54] <AlexLehm> temporalfox: I have a strange issue with the dns resolver, I have removed the async resolver in mail-client and now i get a failure when resolving LOCALHOST

[23:05:06] <skullcrasher> temporalfox, all right, thx

[23:05:11] <temporalfox> do you have an /etc/hosts file ?

[23:05:18] <temporalfox> normally it gets it from there

[23:05:18] <AlexLehm> might be a netty issue though

[23:05:28] <AlexLehm> it works with localhost in lower case

[23:05:33] <temporalfox> what is the failure you are having ?

[23:05:35] <temporalfox> ah

[23:05:37] <temporalfox> never tried that

[23:05:44] <AlexLehm> i am on windows, i think there is localhost in hosts

[23:05:55] <temporalfox> Netty resolver parses the hosts file

[23:05:57] <temporalfox> even on windows

[23:06:31] <AlexLehm> ok, localhost is present in the windows host file

[23:06:50] <AlexLehm> could be a windows issue, i can try my test on a linux

[23:07:48] <skullcrasher> should I post feature wishes/proposals in the github issues project or the google group?

[23:08:19] <temporalfox> if you have a precise idea I would suggest github

[23:08:30] <temporalfox> if you want to discuss, better use the google group

[23:09:09] <skullcrasher> ok. better google group then. It's about support for easy inter cluster communication.

[23:09:55] <AlexLehm> ok, that looks like an issue in netty, when I add LOCALHOST to the hosts file it works

[23:11:19] <temporalfox> skullcrasher looks interesting, eager to read what you suggest

[23:11:45] <temporalfox> AlexLehm what is the expected behavior for DNS resolving ?

[23:11:47] <skullcrasher> temporalfox, well not only suggest but also ask :D

[23:13:01] <AlexLehm> localhost will not always resolve in dns, but it should be present in /etc/hosts

[23:13:20] <AlexLehm> when I use the java classes for the same test, it resolves localhost, LOCALHOST and Localhost all to 127.0.0.1

[23:14:36] <AlexLehm> so localhost is not resolved with a real dns query i assume

[23:15:17] <AlexLehm> which is actually defined by /etc/nsswitch.conf in Linux

[23:15:53] <temporalfox> it would be interesting to know when the LOCALHOST/localhost is done

[23:16:09] <temporalfox> is it done by the JVM when it look at hosts file or not

[23:16:21] <temporalfox> should it be handled by Netty when it resolves against hosts file ?

[23:16:34] <AlexLehm> yes i think so

[23:16:35] <temporalfox> where is defined this precise behavior ?

[23:16:52] <temporalfox> I think we can contribute this to Netty if we can prove it's the correct behavior :-)

[23:17:02] <temporalfox> worse case I think we can patch it in vertx core

[23:17:10] <temporalfox> the resolver is quite flexible

[23:18:08] <AlexLehm> usually a program will use a dnsclient in glibc, which takes a look at the nsswitch file and then /etc/hosts and dns client (and maybe nscd if it is installed)

[23:19:05] <AlexLehm> not sure what java uses, probably they have the same behaviour implement in java directly, i know there are additional cache settings for dns in java, so its probably not using the glibc functions

[23:20:22] <AlexLehm> i think the chances of someone using uppercase LOCALHOST are pretty slim, I used that because I was testing my hostname cert matcher code before, this is not necessary anymore since it now uses the jdk code

[23:40:30] <AlexLehm> ok, that is wrong in the netty code, they use a hostname with case as key

[23:40:31] <AlexLehm> https://github.com/netty/netty/blob/4.1/resolver/src/main/java/io/netty/resolver/HostsFileParser.java#L151

[23:40:46] <AlexLehm> i can create an issue for that