[GitHub] [geode] dschneider-pivotal commented on a change in pull request #5237: cleaned up ExecutionHandlerContext dependencies

[GitBox <gi...@apache.org>]

dschneider-pivotal commented on a change in pull request #5237:
URL: https://github.com/apache/geode/pull/5237#discussion_r439030672



##########
File path: geode-redis/src/main/java/org/apache/geode/redis/internal/netty/ExecutionHandlerContext.java
##########
@@ -264,7 +238,7 @@ private void moveSubscribeToNewEventLoopGroup(ChannelHandlerContext ctx, Command
     if (command.isOfType(SUBSCRIBE)) {
       CountDownLatch latch = new CountDownLatch(0);
       ctx.channel().deregister().addListener((ChannelFutureListener) future -> {
-        subscriberEventLoopGroup.register(ctx.channel()).sync();
+        server.getSubscriberGroup().register(ctx.channel()).sync();

Review comment:
       What I would like to see here is that this code (moveSubscribeToNewEventLoopGroup) to the SubscribeExecutor. Is that possible? I don't really understand what this is about and maybe it is important to do it after the executor has completed.
   I'd also like to see all the netty specific code get moved out of GeodeRedisServer into something like NettyServer in our netty package just to have all the interactions with netty be in that one package. So then this code would interact with the NettyServer for this registration. But then this code would need to get access to the NettyServer (probably from the GeodeRedisServer). If we could break this class's dependency on GeodeRedisServer that might be a good thing but for now we need the GeodeRedisServer anyway (I think so we can ask it to shutdown and if unsupported commands are allowed).




----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org