zhengchenyu created HDFS-15649:
----------------------------------
Summary: the standby namenode's ReplQueues need to keep pace with
active namenode.
Key: HDFS-15649
URL: https://issues.apache.org/jira/browse/HDFS-15649
Project: Hadoop HDFS
Issue Type: Bug
Affects Versions: 3.2.1, 2.7.3
Reporter: zhengchenyu
Fix For: 3.3.1
I think the standby namenode's ReplQueues need to keep pace with active
namenode. You will code in fuction addStoredBlock like below:
{code}
// do not try to handle extra/low redundancy blocks during first safe mode
if (!isPopulatingReplQueues()) {
return storedBlock;
}
{code}
Here, for standby namenode, through I think there are no need to tell standby
to replicate blocks, but need to update neededReconstruction. Because some
metrics need it. For example, missing blocks.
Why I advise this? In our internal version, some bug trigger huge missing block
number. In fact, these blocks are not missing, but addStoredBlock doesn't
update blocks, so keep huge missing block number.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-dev-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-dev-h...@hadoop.apache.org
