[
https://issues.apache.org/jira/browse/HDFS-9498?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15083943#comment-15083943
]
Mingliang Liu commented on HDFS-9498:
-------------------------------------
Thanks for the further discussion.
I should wrote my last comment like this:
{quote}
Suppose the NN is in manual safe mode *and there is no blocks with future GS*,
blockManager.leaveSafeMode(force) will not be able to start the secret manager
*by calling {{startSecretManagerIfNecessary()}}*.
{quote}
The reason is that, in {{startSecretManagerIfNecessary()}}, it will check the
{{isInSafeMode()}} first. This will fail because the manual safe mode is not
reset yet. So after we clear the manual and resource low safe mode, we may need
call startSecretManagerIfNecessary again.
{code}
+ if (blockManager.leaveSafeMode(force)) { // block manager will not start
secret manager successfully in case of manual safe mode
+ setManualAndResourceLowSafeMode(false, false);
+ startSecretManagerIfNecessary(); // call it again after we clear
manual and resource low safe mode.
+ }
{code}
For cases when block manager fails to leave safe mode because of future blocks,
the code is fine without calling {{startSecretManagerIfNecessary}} as it will
fail to start secret manager anyway.
You're right the {{startSecretManagerIfNecessary}} is now protected by NS lock
in all callers. Perhaps adding an assertion {{assert hasWriteLock()}} will be
helpful to keep it being mi-used?
> Move code that tracks orphan blocks to BlockManagerSafeMode
> -----------------------------------------------------------
>
> Key: HDFS-9498
> URL: https://issues.apache.org/jira/browse/HDFS-9498
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: namenode
> Reporter: Mingliang Liu
> Assignee: Mingliang Liu
> Attachments: HDFS-9498.000.patch, HDFS-9498.001.patch,
> HDFS-9498.002.patch, HDFS-9498.003.patch
>
>
> [HDFS-4015] counts and reports orphaned blocks
> {{numberOfBytesInFutureBlocks}} in safe mode. It was implemented in
> {{BlockManager}}. Per discussion in [HDFS-9129] which introduces the
> {{BlockManagerSafeMode}}, we can move code that maintaining orphaned blocks
> to this class.
> Leaving safe mode checks blocks with future GS in {{FSNamesystem}}. This code
> can also be moved to {{BlockManagerSafeMode}}.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
