helix-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Kanak Biscuitwala (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (HELIX-150) Auto rebalance might not evenly distribute states across nodes
Date Mon, 29 Jul 2013 20:47:48 GMT

     [ https://issues.apache.org/jira/browse/HELIX-150?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel

Kanak Biscuitwala updated HELIX-150:

    Attachment: 0001-HELIX-150-Use-partition-global-replica-ids-rather-th.patch

The fix in this case is to assign a unique replica id to each replica of a partition, regardless
of the state. For example, before when we had 1 master and 2 slaves, the replicas would be
(P, M, 0), (P, S, 0), and (P, S, 1). Now they are (P, 0), (P, 1), and (P, 2). There is some
additional cleanup work that I'll file as a separate Jira.

The patch is attached. Code review: https://reviews.apache.org/r/13024/
> Auto rebalance might not evenly distribute states across nodes
> --------------------------------------------------------------
>                 Key: HELIX-150
>                 URL: https://issues.apache.org/jira/browse/HELIX-150
>             Project: Apache Helix
>          Issue Type: Bug
>          Components: helix-core
>            Reporter: Kanak Biscuitwala
>            Assignee: Kanak Biscuitwala
>            Priority: Minor
>         Attachments: 0001-HELIX-150-Use-partition-global-replica-ids-rather-th.patch
> Depending on how nodes come up, some nodes may host too many replicas of a given state,
e.g. a node may end up serving many master replicas. There is also some potential issues as
replicas transition states. It makes more sense just to number the replicas directly instead
of numbering them as (state, state-replica) 

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: http://www.atlassian.com/software/jira

View raw message