cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Thomas Steinmaurer (Jira)" <j...@apache.org>
Subject [jira] [Commented] (CASSANDRA-11418) Nodetool status should reflect hibernate/replacing states
Date Tue, 03 Aug 2021 06:17:00 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-11418?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17392008#comment-17392008
] 

Thomas Steinmaurer commented on CASSANDRA-11418:
------------------------------------------------

As 4.0 has been released, is this something which could be picked up in the near future, perhaps
even for the 3.11 series?

Reason is that showing a node in normal state when using replace_address is not only confusing
for operators, but especially for any automation/monitoring tooling behind a Cassandra cluster.

Additionally, when a Cassandra process disables Gossip (and client protocols) due to disk
issues, other nodes will see it as DN but running nodetool on this particular node will report
UN, although Gossip is disabled. This is additionally confusing for any automation/monitoring
tooling.

> Nodetool status should reflect hibernate/replacing states
> ---------------------------------------------------------
>
>                 Key: CASSANDRA-11418
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-11418
>             Project: Cassandra
>          Issue Type: Improvement
>          Components: Legacy/Observability, Tool/nodetool
>            Reporter: Joel Knighton
>            Assignee: Shaurya Gupta
>            Priority: Low
>              Labels: lhf
>             Fix For: 4.x
>
>         Attachments: cassandra-11418-trunk
>
>
> Currently, the four options for state in nodetool status are joining/leaving/moving/normal.
> Joining nodes are determined based on bootstrap tokens, leaving nodes are based on leaving
endpoints in TokenMetadata, moving nodes are based on moving endpoints in TokenMetadata.
> This means that a node will appear in normal state when going through a bootstrap with
flag replace_address, which can be confusing to operators.
> We should add another state for hibernation/replacing to make this visible. This will
require a way to get a list of all hibernating endpoints.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@cassandra.apache.org
For additional commands, e-mail: commits-help@cassandra.apache.org


Mime
View raw message