[
https://issues.apache.org/jira/browse/ACCUMULO-4640?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16031267#comment-16031267
]
Michael Miller edited comment on ACCUMULO-4640 at 5/31/17 6:06 PM:
-------------------------------------------------------------------
bq. Essentially with "-z" (or "-zi") you are giving the shell the information
that it would otherwise have to figure out from HDFS after looking up the value
of instance.volumes in accumulo-site.xml (or client.conf since 1.8.1).
Correct, except the Shell first looks for "instance.name" and
"instance.zookeeper.host" in client.conf. If the Shell can't find either of
those two properties, it will than try to read them from accumulo-site.xml.
The "instance.volumes" property is not a client property, it is only for use in
the server configuration file, accumulo-site.xml
bq. I guess the question is then: when "-z" (or "-zi") are not provided as
arguments to it should the shell pick up the value of instance.volumes from
accumulo-site.xml or client.conf? 1.8.1 takes the value from client.conf where
1.7.x and 1.8.0 take the value from accumulo-site.xml.
Your client.conf should set the "instance.name" and "instance.zookeeper.host"
properties. This will prevent the shell from having to read accumulo-site.xml.
I can see how this appears to be a bug but this change was made so the Shell
would function as intended (more like a client). Previously it was always
reading from accumulo-site.xml. The Shell was only supposed to fall back to
reading from accumulo-site.xml (for backwards compatibility) when it couldn't
get the information it needed from client.cont. Now, with this change you can
properly secure accumulo-site.xml and run the shell with "-z" (or "-zi" and
"-zh").
was (Author: milleruntime):
bq. Essentially with "-z" (or "-zi") you are giving the shell the information
that it would otherwise have to figure out from HDFS after looking up the value
of instance.volumes in accumulo-site.xml (or client.conf since 1.8.1).
Correct, except the Shell first looks for "instance.name" and
"instance.zookeeper.host" from client.conf. If it can't find either of those
two properties, it will than try to read them from accumulo-site.xml. The
"instance.volumes" property is not a client property, it is only for use in the
server configuration file, accumulo-site.xml
bq. I guess the question is then: when "-z" (or "-zi") are not provided as
arguments to it should the shell pick up the value of instance.volumes from
accumulo-site.xml or client.conf? 1.8.1 takes the value from client.conf where
1.7.x and 1.8.0 take the value from accumulo-site.xml.
Your client.conf should set the "instance.name" and "instance.zookeeper.host"
properties. This will prevent the shell from having to read accumulo-site.xml.
I can see how this appears to be a bug but this change was made so the Shell
would function as intended (more like a client). Previously it was always
reading from accumulo-site.xml. The Shell was only supposed to fall back to
reading from accumulo-site.xml (for backwards compatibility) when it couldn't
get the information it needed from client.cont. Now, with this change you can
properly secure accumulo-site.xml and run the shell with "-z" (or "-zi" and
"-zh").
> Accumulo shell is expecting instance.volumes in client.conf instead of
> accumulo-site.xml
> ----------------------------------------------------------------------------------------
>
> Key: ACCUMULO-4640
> URL: https://issues.apache.org/jira/browse/ACCUMULO-4640
> Project: Accumulo
> Issue Type: Bug
> Components: shell
> Affects Versions: 1.8.1
> Environment: [accumulo@master ~]$ cat /etc/redhat-release
> CentOS Linux release 7.3.1611 (Core)
> [accumulo@master ~]$ uname -a
> Linux master 3.10.0-514.10.2.el7.x86_64 #1 SMP Fri Mar 3 00:04:05 UTC 2017
> x86_64 x86_64 x86_64 GNU/Linux
> [accumulo@master ~]$ java -version
> java version "1.8.0_112"
> Java(TM) SE Runtime Environment (build 1.8.0_112-b15)
> Java HotSpot(TM) 64-Bit Server VM (build 25.112-b15, mixed mode)
> Hadoop: 2.8.0
> ZooKeeper: 3.4.10
> Accumulo: 1.8.1
> Reporter: Luigi Di Fraia
> Fix For: 1.8.2, 2.0.0
>
> Attachments: accumulo-env.sh, accumulo-site.xml
>
>
> Starting with Accumulo 1.8.1 we are getting the exception as per below when
> the Accumulo shell is invoked:
> [accumulo@master ~]$ /usr/local/accumulo/bin/accumulo shell -u root -p
> ***removed***
> 2017-05-18 14:31:23,667 [zookeeper.ZooUtil] ERROR: unable obtain instance id
> at hdfs://master:9000/accumulo/instance_id
> 2017-05-18 14:31:23,669 [start.Main] ERROR: Thread 'shell' died.
> java.lang.RuntimeException: Accumulo not initialized, there is no instance id
> at hdfs://master:9000/accumulo/instance_id
> at
> org.apache.accumulo.core.zookeeper.ZooUtil.getInstanceIDFromHdfs(ZooUtil.java:66)
> at
> org.apache.accumulo.core.zookeeper.ZooUtil.getInstanceIDFromHdfs(ZooUtil.java:51)
> at org.apache.accumulo.shell.Shell.getZooInstance(Shell.java:511)
> at org.apache.accumulo.shell.Shell.setInstance(Shell.java:473)
> at org.apache.accumulo.shell.Shell.config(Shell.java:322)
> at org.apache.accumulo.shell.Shell.execute(Shell.java:589)
> at org.apache.accumulo.start.Main$1.run(Main.java:120)
> at java.lang.Thread.run(Thread.java:745)
> The location used to look up the instance_id
> (hdfs://master:9000/accumulo/instance_id) does not appear to be consistent
> with this site configuration and I am not quite sure how it is being worked
> out.
> The Accumulo Overview web page shows that the instance exists:
> Instance ID: 4264f11a-704f-4c1f-880f-08f6154232b2
> Thu May 18 14:44:03 BST 2017
> The instance_id file is in Hadoop's HDFS:
> [accumulo@master ~]$ /usr/local/hadoop/bin/hdfs dfs -ls
> /user/accumulo/accumulo/instance_id
> Found 1 items
> -rw-r--r-- 3 accumulo supergroup 0 2017-05-18 14:27
> /user/accumulo/accumulo/instance_id/4264f11a-704f-4c1f-880f-08f6154232b2
> Again, within the same environment as selected above apart for using Accumulo
> 1.8.0, the shell can be invoked without issues.
> Here's a snippet from accumulo-site.xml:
> <property>
> <name>instance.volumes</name>
> <value>hdfs://master:9000/user/accumulo/accumulo</value>
> <description>comma separated list of URIs for volumes. example:
> hdfs://localhost:9000/accumulo</description>
> </property>
> For historical reasons our "root volume" is not / but /user/accumulo although
> the above stack trace suggests that / is being used as root volume.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
