Differences
This shows you the differences between two versions of the page.
— |
irc:1459461600 [2017/05/27 13:44] (current) |
||
---|---|---|---|
Line 1: | Line 1: | ||
+ | [03:49:10] <joson> I found that ClusterEventBus keeps all the opened connections connected until the destinated server isn't alive or reachable. | ||
+ | |||
+ | [03:52:13] <joson> Will it consume all the connection resource if the cluster grow up? | ||
+ | |||
+ | [10:28:18] <joson> help | ||
+ | |||
+ | [11:00:28] <cescoffier> joson : it's how the eventbus is working. So it will open connections to all the nodes. What is the size of your cluster (planned ?) | ||
+ | |||
+ | [13:27:26] <Horst> anybody in the mode for a quick question? | ||
+ | |||
+ | [13:27:40] <Horst> I am running into issues uising vertx in ha mode | ||
+ | |||
+ | [13:28:04] <Horst> are some sorts of classloading issues to be known? | ||
+ | |||
+ | [13:28:39] <Horst> i am deploying two verticles in ha mode (same group) | ||
+ | |||
+ | [13:29:04] <Horst> after killing one it get deployed on the other node, so far so good | ||
+ | |||
+ | [13:30:26] <Horst> but I am getting a ClassCastException at a place where a Object is used which is handled by a default codec | ||
+ | |||
+ | [13:33:54] <Horst> this makes me wonder if I am runing into classloading issues. Same class but loaded by different classloaders | ||
+ | |||
+ | [14:29:34] <cescoffier> Horst : HA requires that the code that is redeployed is "available" on the node where it will be redeployed. Available means in the classpath (or whatever classloading strategy you use) | ||
+ | |||
+ | [14:29:57] <cescoffier> if it's the case, it may be a bug, and in this case it would be amazing if you an produce a reproducer | ||
+ | |||
+ | [15:00:36] <Horst> I think I understand whats happening | ||
+ | |||
+ | [15:01:57] <Horst> both verticles are started with the vertx command using the -cp switch like so "vertx run myVerticle -cluster -conf '{\"name\":\"mst-ha1\"}' -cp ./libs/my.jar -cluster-host 192.168.3.45 -ha -hagroup g1" | ||
+ | |||
+ | [15:03:06] <Horst> actually I deploy the same verticle twice but with different configurations | ||
+ | |||
+ | [15:03:48] <Horst> killing one off them redeploys it on the other. This is working! :-) | ||
+ | |||
+ | [15:04:53] <Horst> but the -cp switch leads to the creation of a new Classloader in the io.vertx.core.impl.launcher.commands.ClasspathHandler | ||
+ | |||
+ | [15:06:40] <Horst> so that code like this " private void sendMessage(Message<MyMessage> msg) {...}" leads to a ClassCastException. Same Class but different Classloaders | ||
+ | |||
+ | [15:09:25] <Horst> I can circumvent it right now by omitting the "-cp" switch and placing my jar into the %VERTX_HOME%\lib\ directory | ||
+ | |||
+ | [15:10:09] <Horst> any other ideas? | ||
+ | |||
+ | [15:11:30] <Horst> MyMessage btw is handled by a DefaultCodec | ||
+ | |||
+ | [17:21:02] <kuwv> Do I have to setup JRUBY_HOME to use ruby scripts in vertx? | ||
+ | |||
+ | [21:49:01] <kuwv> Is anyone on right now?' | ||
+ | |||
+ | [22:37:02] <kuwv> Hello | ||