ignite-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Semen Boikov (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (IGNITE-1229) Stop ping procedure when node left event received
Date Wed, 12 Aug 2015 07:44:45 GMT

    [ https://issues.apache.org/jira/browse/IGNITE-1229?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14693069#comment-14693069
] 

Semen Boikov commented on IGNITE-1229:
--------------------------------------

My comments:
- ring.node() still can return null event after nodeAlive returned true
- I don't think it makes sence to add special IgniteNodeLeftException, instead method 'IgniteBiTuple<UUID,
Boolean> pingNode(InetSocketAddress addr, UUID nodeId, UUID clientNodeId)' can just return
'null'
- method dumpDebugInfo already synchronizes on 'mux' so your change in this method isn't needed
- it would be good also add interruption for case when pinging thread sits inside 'sock.connect'
and at this moment node failure is detected

> Stop ping procedure when node left event received 
> --------------------------------------------------
>
>                 Key: IGNITE-1229
>                 URL: https://issues.apache.org/jira/browse/IGNITE-1229
>             Project: Ignite
>          Issue Type: Task
>          Components: general
>            Reporter: Semen Boikov
>            Assignee: Denis Magda
>             Fix For: ignite-1.4
>
>         Attachments: ignite-1229.patch
>
>
> Node ping procedure (ServerImpl.pingNode(TcpDiscoveryNode node)) can be time consuming:
by default it 10 times tries to connect to each node's address, and now it continues to ping
even after node left/failed event is received.
> Also there is minor issue in ServerImpl.pingNode(UUID nodeId): if ping returns 'false'
then before initiating status check need to check if node is still visible.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message