[
https://issues.apache.org/jira/browse/HDFS-12521?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16207099#comment-16207099
]
Yiqun Lin edited comment on HDFS-12521 at 10/17/17 9:14 AM:
------------------------------------------------------------
Thanks for working on this, [~ljain]. I'd like to take the first review, :).
The following are my review comments (just some minor comments):
*ContainerMapping.java*:
line105:looks like '+' is not needed.
line148:TODO comments should be updated.
*ContainerStateManager.java*
line204: We can just print the error info instead of checking message then log
error. The error message contained in exception can be showed in logging.
Another point, please use {{LOG.error}} here.
line397:Seems we can reach max size of container, change {{info.getAllocated()
+ size < this.containerSize}} to {{info.getAllocated() + size <=
this.containerSize}}
line429:Please use {{LOG.error}} to log error.
bq.MetadataStoreBuilder should use caching provided by LevelDb and RocksDb for
faster db operations. LevelDb and RocksDb are internally configured to use a
default cache of size ...
I am a little confused for this. Now we have already one config named
{{ozone.scm.db.cache.size.mb}} to pass to MetadataStoreBuilder which can avoid
disk I/O. Can you explain the difference with what you mentioned? Maybe I am
missing something.
>From the Jenkins report, current patch can't compiled, could you please take a
>look of this? Also please fix current warnings in your next patch. Thanks.
was (Author: linyiqun):
Thanks for working on this, [~ljain]. I'd like to take the first review, :).
The following are my review comments (just some minor comments):
*ContainerMapping.java*:
line105:looks like '+' is not needed.
line148:TODO comments should be updated.
*ContainerStateManager.java*
line204: We can just print the error info instead of checking message then log
error. The error message contained in exception can be showed in logging.
Another point, please use {{LOG.error}} here.
line397:Seems we can reach max size of container, change {{info.getAllocated()
+ size < this.containerSize }} to {{info.getAllocated() + size <=
this.containerSize }}
line429:Please use {{LOG.error}} to log error.
bq.MetadataStoreBuilder should use caching provided by LevelDb and RocksDb for
faster db operations. LevelDb and RocksDb are internally configured to use a
default cache of size ...
I am a little confused for this. Now we have already one config named
{{ozone.scm.db.cache.size.mb}} to pass to MetadataStoreBuilder which can avoid
disk I/O. Can you explain the difference with what you mentioned? Maybe I am
missing something.
>From the Jenkins report, current patch can't compiled, could you please take a
>look of this? Also please fix current warnings in your next patch. Thanks.
> Ozone: SCM should read all Container info into memory when booting up
> ---------------------------------------------------------------------
>
> Key: HDFS-12521
> URL: https://issues.apache.org/jira/browse/HDFS-12521
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: ozone
> Reporter: Anu Engineer
> Assignee: Lokesh Jain
> Labels: ozoneMerge
> Attachments: HDFS-12521-HDFS-7240.001.patch,
> HDFS-12521-HDFS-7240.002.patch
>
>
> When SCM boots up it should read all containers into memory. This is a
> performance optimization that allows delays on SCM side. This JIRA tracks
> that issue.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
