[ 
https://issues.apache.org/jira/browse/HDDS-10326?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17817431#comment-17817431
 ] 

Nandakumar commented on HDDS-10326:
-----------------------------------

[~arp], as you mentioned the problem is with {{ozone.metadata.dirs}} 
configuration which is used by OM, SCM, and the Datanode.

If a separate {{Ratis}} directory is not configured for OM and SCM, they fall 
back to {{ozone.metadata.dirs}} location

SCM uses  <<{{{}ozone.metadata.dirs{}}}>>/scm-ha
OM uses  <<{{{}ozone.metadata.dirs{}}}>>/ratis
Datanode uses  <<{{{}ozone.metadata.dirs{}}}>>/ratis

Since OM and Datanode use the same location, configuring OM and Datanode on the 
same node will not work unless we specify separate {{ratis}} dir for OM.

We should ideally use <<{{{}ozone.metadata.dirs{}}}>>/scm, 
<<{{{}ozone.metadata.dirs{}}}>>/om and <<{{{}ozone.metadata.dirs{}}}>>/datanode 
for SCM, OM, and Datanodes respectively, but we cannot make that change now 
without breaking upgrades.

+1 on fixing the document.

> Ozone single-node cluster cannot be started by following the docs
> -----------------------------------------------------------------
>
>                 Key: HDDS-10326
>                 URL: https://issues.apache.org/jira/browse/HDDS-10326
>             Project: Apache Ozone
>          Issue Type: Bug
>          Components: documentation
>            Reporter: Arpit Agarwal
>            Assignee: Sarveksha Yeshavantha Raju
>            Priority: Major
>              Labels: newbie
>
> It is not possible to start an Ozone single-node cluster using the 
> instructions on the cwiki here: 
> https://cwiki.apache.org/confluence/display/OZONE/Single+Node+Deployment
> Using the {{ozone genconf}} doesn't work either. The only way I could get it 
> to work is by specifying separate config dirs for OM, SCM and Datanode that 
> have different values for {{ozone.metadata.dirs}}. I didn't investigate in 
> depth however the problem seems to be that DataNode uses the same metadata 
> dir as the SCM and cannot get the directory lock (thankfully we have that 
> safeguard).
> We should at the very least fix the documentation with correct instructions, 
> and ideally make it simpler to start a single node cluster by choosing 
> sensible defaults.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to