[
https://issues.apache.org/jira/browse/HDFS-14208?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
xuzq updated HDFS-14208:
------------------------
Description:
When cluster is very large, Standby startup takes about 1 hours. After Standby
starts and exits SafeMode, a large number of missingblocks often appeared after
failover from standby to active. And the missingblocks will gradually
disappear after 6 hours(block report interval).
According the log and find only half the blocks of one DataNode be processed in
SaftMode.
Because the lease expired after processed some blocks of one DataNode.
[MergeRequest|https://github.com/apache/hadoop/pull/467]
was:
When cluster is very large, Standby startup takes about 1 hours. After Standby
starts and exits SafeMode, a large number of missingblocks often appeared after
failover from standby to active. And the missingblocks will gradually
disappear after 6 hours(block report interval).
According the log and find only half the blocks of one DataNode be processed in
SaftMode.
Because the lease expired after processed some blocks of one DataNode.
> A large number missingblocks happend after failover to active.
> --------------------------------------------------------------
>
> Key: HDFS-14208
> URL: https://issues.apache.org/jira/browse/HDFS-14208
> Project: Hadoop HDFS
> Issue Type: Improvement
> Components: hdfs
> Affects Versions: 3.0.2, 3.1.1, 3.0.3
> Reporter: xuzq
> Priority: Major
>
> When cluster is very large, Standby startup takes about 1 hours. After
> Standby starts and exits SafeMode, a large number of missingblocks often
> appeared after failover from standby to active. And the missingblocks will
> gradually disappear after 6 hours(block report interval).
> According the log and find only half the blocks of one DataNode be processed
> in SaftMode.
>
> Because the lease expired after processed some blocks of one DataNode.
> [MergeRequest|https://github.com/apache/hadoop/pull/467]
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
