[jira] [Commented] (IGNITE-5103) TcpDiscoverySpi ignores maxMissedClientHeartbeats property

["Andrew Mashenkov (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/IGNITE-5103?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16191327#comment-16191327 ] 

Andrew Mashenkov commented on IGNITE-5103:
------------------------------------------

I'd think that with having maxMissedClientHeartbeats and heartbeatFrequency and failureDetectionTimeout parameters 
grid should drop slow client node if it missed "maxMissedClientHeartbeats" during failureDetectionTimeout.

But looking at a code I see we just calculate failureDetectionTimeout for client nodes as maxMissedClientHeartbeats*heartbeatFrequency.

From user perspective, having failureDetectionTimeout and clientFailureDetectionTimeout is much less confusing.
So, looks like this can be closed as won't fix.



> TcpDiscoverySpi ignores maxMissedClientHeartbeats property
> ----------------------------------------------------------
>
>                 Key: IGNITE-5103
>                 URL: https://issues.apache.org/jira/browse/IGNITE-5103
>             Project: Ignite
>          Issue Type: Bug
>          Components: general
>    Affects Versions: 1.9
>            Reporter: Valentin Kulichenko
>            Assignee: Dmitry Karachentsev
>             Fix For: 2.3
>
>         Attachments: TcpDiscoveryClientSuspensionSelfTest.java
>
>
> Test scenario is the following:
> * Start one or more servers.
> * Start a client node.
> * Suspend client process using {{-SIGSTOP}} signal.
> * Wait for {{maxMissedClientHeartbeats*heartbeatFrequency}}.
> * Client node is expected to be removed from topology, but server nodes don't do that.
> Attached is the unit test reproducing the same by stopping the heartbeat sender thread on the client.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)