cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Sylvain Lebresne (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (CASSANDRA-10809) Create a -D option to prevent gossip startup
Date Thu, 17 Mar 2016 16:23:33 GMT

     [ https://issues.apache.org/jira/browse/CASSANDRA-10809?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Sylvain Lebresne updated CASSANDRA-10809:
-----------------------------------------
    Attachment: 10809.txt

I believe restoring the previous behavior is as simple as aborting {{initServer}} before {{prepareToJoin}},
which the attached patch does. The patch doesn't add a way to start gossip after the fact,
but I assume that if you're debugging a schema problem or something like that, it's ok to
restart (without the flag) once you've fixed your problem.

> Create a -D option to prevent gossip startup
> --------------------------------------------
>
>                 Key: CASSANDRA-10809
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-10809
>             Project: Cassandra
>          Issue Type: Improvement
>          Components: Distributed Metadata
>            Reporter: Brandon Williams
>             Fix For: 2.1.x
>
>         Attachments: 10809.txt
>
>
> In CASSANDRA-6961 we changed how join_ring=false works, to great benefit.  However, sometimes
you need to a node to come up, but not interact with other nodes whatsoever - for example
if you have a schema problem, it will still pull the schema from another node because they
still gossip even though we're in a dead state.
> We can add a way to restore the previous behavior by simply adding something like -Dcassandra.start_gossip=false.
> In the meantime we can workaround by setting listen_address to localhost, but that's
kind of a pain.



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

Mime
View raw message