You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@kudu.apache.org by "Todd Lipcon (JIRA)" <ji...@apache.org> on 2017/12/26 15:55:00 UTC
[jira] [Commented] (KUDU-2239) Kudu Maste Rpc ListTabletServers
Have Dead Tablet Server
[ https://issues.apache.org/jira/browse/KUDU-2239?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16303908#comment-16303908 ]
Todd Lipcon commented on KUDU-2239:
-----------------------------------
The alive/dead status is currently just based on millis_since_heartbeat, which is already returned in the response. You can filter out any servers which have not heartbeat in 60 seconds (or whatever other threshold you choose). Is that not sufficient?
> Kudu Maste Rpc ListTabletServers Have Dead Tablet Server
> ---------------------------------------------------------
>
> Key: KUDU-2239
> URL: https://issues.apache.org/jira/browse/KUDU-2239
> Project: Kudu
> Issue Type: Bug
> Components: api
> Affects Versions: 1.5.0
> Reporter: yangz
>
> It is the rpc request of the masterserviceimpl
> {code:java}
> virtual void ListTabletServers(const ListTabletServersRequestPB* req,
> ListTabletServersResponsePB* resp,
> rpc::RpcContext* rpc) OVERRIDE;
> {code}
> The rpc return all the tablet server, no matter it's live or dead. And the response have no description of the tablet server.
> When I want to collect the metrics from all the tserver, I invoke the ListTabletServers rpc.
> Then I send all the request to the tablet server to get the metrics. But some tablet server is dead.
> I think this rpc will always return the live tablet server, or add some status of the tablet server, or
> add a rpc to return the live tablet servers.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)