You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@cassandra.apache.org by "Jonathan Ellis (JIRA)" <ji...@apache.org> on 2010/01/04 17:19:58 UTC
[jira] Created: (CASSANDRA-665) Persist token ring across cluster
restarts
Persist token ring across cluster restarts
------------------------------------------
Key: CASSANDRA-665
URL: https://issues.apache.org/jira/browse/CASSANDRA-665
Project: Cassandra
Issue Type: New Feature
Components: Core
Reporter: Jonathan Ellis
Assignee: Jonathan Ellis
Priority: Minor
Fix For: 0.9
This prevents users from shooting their foot by shutting down a cluster, then restarting it with a subset of the former nodes.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (CASSANDRA-665) Persist token ring across cluster
restarts
Posted by "Jonathan Ellis (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/CASSANDRA-665?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12796309#action_12796309 ]
Jonathan Ellis commented on CASSANDRA-665:
------------------------------------------
So a node would add a UUID applicationState that it gossips, and if another node sees that UUID coming from a different IP, it can replace the old IP with the new one in TokenMetadata etc. (I think we want to keep using IP addresses internally for e.g. MessagingService rather than doing an extra lookup on each message send.)
> Persist token ring across cluster restarts
> ------------------------------------------
>
> Key: CASSANDRA-665
> URL: https://issues.apache.org/jira/browse/CASSANDRA-665
> Project: Cassandra
> Issue Type: New Feature
> Components: Core
> Reporter: Jonathan Ellis
> Priority: Minor
> Fix For: 0.9
>
>
> This prevents users from shooting their foot by shutting down a cluster, then restarting it with a subset of the former nodes.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Updated: (CASSANDRA-665) Persist token ring across cluster
restarts
Posted by "Jonathan Ellis (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/CASSANDRA-665?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Jonathan Ellis updated CASSANDRA-665:
-------------------------------------
Assignee: (was: Jonathan Ellis)
> Persist token ring across cluster restarts
> ------------------------------------------
>
> Key: CASSANDRA-665
> URL: https://issues.apache.org/jira/browse/CASSANDRA-665
> Project: Cassandra
> Issue Type: New Feature
> Components: Core
> Reporter: Jonathan Ellis
> Priority: Minor
> Fix For: 0.9
>
>
> This prevents users from shooting their foot by shutting down a cluster, then restarting it with a subset of the former nodes.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Updated: (CASSANDRA-665) Persist token ring across cluster
restarts
Posted by "Jonathan Ellis (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/CASSANDRA-665?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Jonathan Ellis updated CASSANDRA-665:
-------------------------------------
Fix Version/s: (was: 0.6)
0.7
> Persist token ring across cluster restarts
> ------------------------------------------
>
> Key: CASSANDRA-665
> URL: https://issues.apache.org/jira/browse/CASSANDRA-665
> Project: Cassandra
> Issue Type: New Feature
> Components: Core
> Reporter: Jonathan Ellis
> Priority: Minor
> Fix For: 0.7
>
>
> This prevents users from shooting their foot by shutting down a cluster, then restarting it with a subset of the former nodes.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (CASSANDRA-665) Persist token ring across cluster
restarts
Posted by "Kelvin Kakugawa (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/CASSANDRA-665?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12796307#action_12796307 ]
Kelvin Kakugawa commented on CASSANDRA-665:
-------------------------------------------
If you could keep version vectors in mind when you think about UUIDs, I'd appreciate it.
Right now, I'm using the IP address as the node id. However, if a UUID -> IP mapping is implemented, I'll use the UUID. So, if possible, I'd prefer the UUID to be shorter than longer. Especially, since, it's not practical for the cluster to be more than 300 nodes or so. (Unless the gossip implementation is modified to be hierarchical.)
> Persist token ring across cluster restarts
> ------------------------------------------
>
> Key: CASSANDRA-665
> URL: https://issues.apache.org/jira/browse/CASSANDRA-665
> Project: Cassandra
> Issue Type: New Feature
> Components: Core
> Reporter: Jonathan Ellis
> Priority: Minor
> Fix For: 0.9
>
>
> This prevents users from shooting their foot by shutting down a cluster, then restarting it with a subset of the former nodes.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (CASSANDRA-665) Persist token ring across cluster
restarts
Posted by "Kelvin Kakugawa (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/CASSANDRA-665?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12796542#action_12796542 ]
Kelvin Kakugawa commented on CASSANDRA-665:
-------------------------------------------
I feel you. Generating compact UUIDs is not a trivial task. However, this problem particularly impacts me, so I'll keep an eye out for solutions.
> Persist token ring across cluster restarts
> ------------------------------------------
>
> Key: CASSANDRA-665
> URL: https://issues.apache.org/jira/browse/CASSANDRA-665
> Project: Cassandra
> Issue Type: New Feature
> Components: Core
> Reporter: Jonathan Ellis
> Priority: Minor
> Fix For: 0.9
>
>
> This prevents users from shooting their foot by shutting down a cluster, then restarting it with a subset of the former nodes.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Issue Comment Edited: (CASSANDRA-665) Persist token ring
across cluster restarts
Posted by "Jonathan Ellis (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/CASSANDRA-665?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12796295#action_12796295 ]
Jonathan Ellis edited comment on CASSANDRA-665 at 1/4/10 7:29 PM:
------------------------------------------------------------------
The naive solution of just using the IP->token mapping we already persist trades this problem for a new one: it would become difficult to change the IP of an existing node. Granted, this is a rare operation -- but so is shutting down the entire cluster.
One way to address this is using UUIDs for identity (rather than IP, directly) and mapping UUID -> token, UUID -> IP separately.
was (Author: jbellis):
The naive solution of just using the IP->token mapping wil already persist trades this problem for a new one: it would become difficult to change the IP of an existing node. Granted, this is a rare operate -- but so is shutting down the entire cluster.
One way to address this is using UUIDs for identity (rather than IP, directly) and mapping UUID -> token, UUID -> IP separately.
> Persist token ring across cluster restarts
> ------------------------------------------
>
> Key: CASSANDRA-665
> URL: https://issues.apache.org/jira/browse/CASSANDRA-665
> Project: Cassandra
> Issue Type: New Feature
> Components: Core
> Reporter: Jonathan Ellis
> Priority: Minor
> Fix For: 0.9
>
>
> This prevents users from shooting their foot by shutting down a cluster, then restarting it with a subset of the former nodes.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (CASSANDRA-665) Persist token ring across cluster
restarts
Posted by "Jonathan Ellis (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/CASSANDRA-665?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12796520#action_12796520 ]
Jonathan Ellis commented on CASSANDRA-665:
------------------------------------------
If you know of a good way to get unique ids in less than 32 bits (the size of an ipv4 address) then great. But even a MAC address alone is 48 bits so I'm not sure that's a reasonable thing to hope for. :)
> Persist token ring across cluster restarts
> ------------------------------------------
>
> Key: CASSANDRA-665
> URL: https://issues.apache.org/jira/browse/CASSANDRA-665
> Project: Cassandra
> Issue Type: New Feature
> Components: Core
> Reporter: Jonathan Ellis
> Priority: Minor
> Fix For: 0.9
>
>
> This prevents users from shooting their foot by shutting down a cluster, then restarting it with a subset of the former nodes.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (CASSANDRA-665) Persist token ring across cluster
restarts
Posted by "Jonathan Ellis (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/CASSANDRA-665?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12796295#action_12796295 ]
Jonathan Ellis commented on CASSANDRA-665:
------------------------------------------
The naive solution of just using the IP->token mapping wil already persist trades this problem for a new one: it would become difficult to change the IP of an existing node. Granted, this is a rare operate -- but so is shutting down the entire cluster.
One way to address this is using UUIDs for identity (rather than IP, directly) and mapping UUID -> token, UUID -> IP separately.
> Persist token ring across cluster restarts
> ------------------------------------------
>
> Key: CASSANDRA-665
> URL: https://issues.apache.org/jira/browse/CASSANDRA-665
> Project: Cassandra
> Issue Type: New Feature
> Components: Core
> Reporter: Jonathan Ellis
> Assignee: Jonathan Ellis
> Priority: Minor
> Fix For: 0.9
>
>
> This prevents users from shooting their foot by shutting down a cluster, then restarting it with a subset of the former nodes.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.