cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Robert Coli (JIRA)" <>
Subject [jira] [Created] (CASSANDRA-5187) be consistent in visible messages about DOWN versus dead state
Date Fri, 25 Jan 2013 21:15:13 GMT
Robert Coli created CASSANDRA-5187:

             Summary: be consistent in visible messages about DOWN versus dead state
                 Key: CASSANDRA-5187
             Project: Cassandra
          Issue Type: Bug
          Components: Core
    Affects Versions: 1.2.0
            Reporter: Robert Coli
            Priority: Trivial

Cassandra has two states that seem to be used interchangeably in some messages.

1) "DOWN" state, which simply means that the node has not responded in the last Gossip round.
2) "dead" which means a state listed in DEAD_STATES in

static final List<String> DEAD_STATES = Arrays.asList(VersionedValue.REMOVING_TOKEN,
VersionedValue.REMOVED_TOKEN,VersionedValue.STATUS_LEFT, VersionedValue.HIBERNATE);

However, it seems to use the terms incorrectly in the following places :

a) the log message :
""InetAddress {} is now dead.", addr);

this is in Gossiper::markDead (which ideally would also be renamed, because it does not seem
to put anything in one of the DEAD_STATES..)

but in Gossiper::markAlive the paired log message is :
""InetAddress {} is now UP", addr);

If being marked alive means you are UP, then being marked dead should mean you are DOWN.

b) NodeToolHelp.yaml has :
Disable gossip (effectively marking the node dead)

Attached is a patch against trunk which changes both "dead"s to "DOWN".

This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see:

View raw message