cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Mike Heffner (JIRA)" <>
Subject [jira] [Commented] (CASSANDRA-5337) vnode-aware replacenode command
Date Thu, 22 Aug 2013 22:55:52 GMT


Mike Heffner commented on CASSANDRA-5337:

Is it correct that these two conditionals should be updated with the same method:

@@ -574,7 +574,7 @@ public class StorageService extends NotificationBroadcasterSupport implements
         appStates.put(ApplicationState.NET_VERSION, valueFactory.networkVersion());
         appStates.put(ApplicationState.HOST_ID, valueFactory.hostId(SystemTable.getLocalHostId()));
         appStates.put(ApplicationState.RPC_ADDRESS, valueFactory.rpcaddress(DatabaseDescriptor.getRpcAddress()));
-        if (0 != DatabaseDescriptor.getReplaceTokens().size())
+        if (DatabaseDescriptor.isReplacing())


@@ -655,7 +655,7 @@ public class StorageService extends NotificationBroadcasterSupport implements
             if (logger.isDebugEnabled())
                 logger.debug("... got ring + schema info");
-            if (DatabaseDescriptor.getReplaceTokens().size() == 0)
+            if (DatabaseDescriptor.isReplacing())

(another case later in patch)

It would seem like one of those should be {{!DatabaseDescriptor.isReplacing()}}.

Apologies if I'm reading this wrong, just looked wrong on initial read.
> vnode-aware replacenode command
> -------------------------------
>                 Key: CASSANDRA-5337
>                 URL:
>             Project: Cassandra
>          Issue Type: New Feature
>          Components: Core
>    Affects Versions: 1.2.0
>            Reporter: Jonathan Ellis
>            Assignee: Brandon Williams
>              Labels: vnodes
>             Fix For: 1.2.7, 2.0
>         Attachments: 5337.txt, 5337-v2.txt
> Currently you have the following options to replace a dead, unrecoverable node:
> - replacetoken.  this requires specifying all 256 or so vnode tokens as a CSL
> - bootstrap new node, decommission old one.  this is inefficient since the new node's
vnodes will probably not overlap much with the old one's, so we replicate stream about 2x
as much as if we were just replacing the old with the new
> We should add an analogue to replacetoken that takes the address or node ID of the dead
node instead.

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