[
https://issues.apache.org/jira/browse/HDDS-6259?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Stephen O'Donnell updated HDDS-6259:
------------------------------------
Description: The pipeline returned from SCM for EC closed containers and
closed pipelines does not carry the correct replica index with the datanode
list. This means the replicas are not returned in the correct order resulting
in strange errors reading EC data, such as wrong data returned and
"insufficient nodes" errors when there are enough nodes available. (was: With
a small key less than 1 chunk, if the replica with index 1 is up, but more than
parity_num other replicas are not available, attempting to read the key gives
an insufficient datanode errors. Ideally, as index_1 is present, and the
keyLength is less than 1 chunk, the data should be readable.
A similar scenario for a key less than 1 chunk, is that indexes 1, 2, 3 and 4
OR 5 are down. As we know 2 and 3 are only zeros, and we have a single parity,
we should be able to reconstruct this, but again it fails with an insufficient
datanodes error.)
Summary: EC: Pipelines for closed containers should contain correct
replica indexes (was: EC: Small keys under 1 chunk should tolerate more than
parity nodes being unavailable)
> EC: Pipelines for closed containers should contain correct replica indexes
> --------------------------------------------------------------------------
>
> Key: HDDS-6259
> URL: https://issues.apache.org/jira/browse/HDDS-6259
> Project: Apache Ozone
> Issue Type: Sub-task
> Reporter: Stephen O'Donnell
> Assignee: Stephen O'Donnell
> Priority: Major
>
> The pipeline returned from SCM for EC closed containers and closed pipelines
> does not carry the correct replica index with the datanode list. This means
> the replicas are not returned in the correct order resulting in strange
> errors reading EC data, such as wrong data returned and "insufficient nodes"
> errors when there are enough nodes available.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
