[
https://issues.apache.org/jira/browse/HDFS-16653?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jingxuan Fu updated HDFS-16653:
-------------------------------
Description:
{code:java}
<property>
<name>dfs.client.mmap.cache.size</name>
<value>256</value>
<description>
When zero-copy reads are used, the DFSClient keeps a cache of recently used
memory mapped regions. This parameter controls the maximum number of
entries that we will keep in that cache.
The larger this number is, the more file descriptors we will potentially
use for memory-mapped files. mmaped files also use virtual address space.
You may need to increase your ulimit virtual address space limits before
increasing the client mmap cache size.
Note that you can still do zero-copy reads when this size is set to 0.
</description>
</property>
{code}
When the configuration item “dfs.client.mmap.cache.size” is set to a negative
number, it will cause /hadoop/bin hdfs dfsadmin -safemode provides all the
operation options including enter, leave, get, wait and forceExit are invalid,
the terminal returns security mode is null and no exceptions are thrown.
{code:java}
hadoop@ljq1:~/hadoop-3.1.3-work/etc/hadoop$ hdfs dfsadmin -safemode leave
safemode: null
Usage: hdfs dfsadmin [-safemode enter | leave | get | wait | forceExit] {code}
In summary, I think we need to improve the check mechanism related to this
configuration item, add maxEvictableMmapedSize that is
"dfs.client.mmap.cache.size" related Precondition check suite error message，and
give a clear indication when the configuration is abnormal in order to solve
the problem in time and reduce the impact on the safe mode related operations.
was:
{code:java}
<property>
<name>dfs.client.mmap.cache.size</name>
<value>256</value>
<description>
When zero-copy reads are used, the DFSClient keeps a cache of recently used
memory mapped regions. This parameter controls the maximum number of
entries that we will keep in that cache.
The larger this number is, the more file descriptors we will potentially
use for memory-mapped files. mmaped files also use virtual address space.
You may need to increase your ulimit virtual address space limits before
increasing the client mmap cache size.
Note that you can still do zero-copy reads when this size is set to 0.
</description>
</property>
{code}
When the configuration item “dfs.client.mmap.cache.size” is set to a negative
number, it will cause /hadoop/bin hdfs dfsadmin -safemode provides all the
operation options including enter, leave, get, wait and forceExit are invalid,
the terminal returns security mode is null and no exceptions are thrown.
{code:java}
hadoop@ljq1:~/hadoop-3.1.3-work/etc/hadoop$ hdfs dfsadmin -safemode leave
safemode: null
Usage: hdfs dfsadmin [-safemode enter | leave | get | wait | forceExit] {code}
In summary, I think we need to improve the check mechanism related to this
configuration item, and give a clear indication when the configuration is
abnormal in order to solve the problem in time and reduce the impact on the
safe mode related operations.
> Safe mode related operations cannot be performed when
> “dfs.client.mmap.cache.size” is set to a negative number
> --------------------------------------------------------------------------------------------------------------
>
> Key: HDFS-16653
> URL: https://issues.apache.org/jira/browse/HDFS-16653
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: dfsadmin
> Affects Versions: 3.1.3
> Environment: Linux version 4.15.0-142-generic
> (buildd@lgw01-amd64-039) (gcc version 5.4.0 20160609 (Ubuntu
> 5.4.0-6ubuntu1~16.04.12))
> Reporter: Jingxuan Fu
> Assignee: Jingxuan Fu
> Priority: Major
> Labels: pull-request-available
>
>
> {code:java}
> <property>
> <name>dfs.client.mmap.cache.size</name>
> <value>256</value>
> <description>
> When zero-copy reads are used, the DFSClient keeps a cache of recently
> used
> memory mapped regions. This parameter controls the maximum number of
> entries that we will keep in that cache.
> The larger this number is, the more file descriptors we will potentially
> use for memory-mapped files. mmaped files also use virtual address space.
> You may need to increase your ulimit virtual address space limits before
> increasing the client mmap cache size.
>
> Note that you can still do zero-copy reads when this size is set to 0.
> </description>
> </property>
> {code}
> When the configuration item “dfs.client.mmap.cache.size” is set to a negative
> number, it will cause /hadoop/bin hdfs dfsadmin -safemode provides all the
> operation options including enter, leave, get, wait and forceExit are
> invalid, the terminal returns security mode is null and no exceptions are
> thrown.
> {code:java}
> hadoop@ljq1:~/hadoop-3.1.3-work/etc/hadoop$ hdfs dfsadmin -safemode leave
> safemode: null
> Usage: hdfs dfsadmin [-safemode enter | leave | get | wait | forceExit] {code}
> In summary, I think we need to improve the check mechanism related to this
> configuration item, add maxEvictableMmapedSize that is
> "dfs.client.mmap.cache.size" related Precondition check suite error
> message，and give a clear indication when the configuration is abnormal in
> order to solve the problem in time and reduce the impact on the safe mode
> related operations.
>
>
>
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
