[ https://issues.apache.org/jira/browse/SPARK-11825?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Marcelo Vanzin resolved SPARK-11825. ------------------------------------ Resolution: Not A Problem >From [~zsxwing]: {quote} It's a false alarm. `IdleStateHandler` will reset the timer in `channelReadComplete`, which isn't intercepted by us. So `IdleStateHandler` will be notified for each read operation even if the readable bytes contains only a part of frame. {quote} > TransportContext's IdleStateHandler should be the first handler in the > pipeline > ------------------------------------------------------------------------------- > > Key: SPARK-11825 > URL: https://issues.apache.org/jira/browse/SPARK-11825 > Project: Spark > Issue Type: Bug > Components: Spark Core > Affects Versions: 1.6.0 > Reporter: Marcelo Vanzin > Priority: Minor > > TransportContext installs an IdleStateHandler on every channel created, to > implement timeouts when the channel is waiting for data but none seems to > arrive. > But the handler is installed last in the pipeline; so if there's a really > large frame arriving, and it takes a while for it to be transferred, that can > trigger the timeout even though there is data flowing through the channel. -- This message was sent by Atlassian JIRA (v6.3.4#6332) --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org