hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Duo Zhang (Jira)" <j...@apache.org>
Subject [jira] [Resolved] (HBASE-24517) AssignmentManager.start should add meta region to ServerStateNode
Date Mon, 15 Jun 2020 02:47:00 GMT

     [ https://issues.apache.org/jira/browse/HBASE-24517?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Duo Zhang resolved HBASE-24517.
-------------------------------
    Resolution: Fixed

> AssignmentManager.start should add meta region to ServerStateNode
> -----------------------------------------------------------------
>
>                 Key: HBASE-24517
>                 URL: https://issues.apache.org/jira/browse/HBASE-24517
>             Project: HBase
>          Issue Type: Bug
>          Components: amv2
>            Reporter: Duo Zhang
>            Assignee: Duo Zhang
>            Priority: Critical
>             Fix For: 3.0.0-alpha-1, 2.3.0, 2.2.6
>
>
> In AssignmentManager.start, we will load the meta region state and location from zk and
create the RegionStateNode, but we forget to call regionStates.addRegionToServer to add the
region to the region server.
> Found this when implementing HBASE-24390. As in HBASE-24390, we will remove RegionInfoBuilder.FIRST_META_REGIONINFO
so in SCP, we need to use the getRegionsOnServer instead of RegionInfoBuilder.FIRST_META_REGIONINFO
when assigning meta, so the bug becomes a real problem.
> Though it is not a big problem for SCP for current 2.x and master branches, it is a high
risky bug. For example, in AssignmentManager.submitServerCrash, now we use the RegionStateNode
of meta regions to determine whether the given region server carries meta regions. But it
is also valid to test through the ServerStateNode's region list. If later we change this method
to use ServerStateNode, it will cause very serious data loss bug.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

Mime
View raw message