errose28 commented on PR #3948:
URL: https://github.com/apache/ozone/pull/3948#issuecomment-1320639167

   > I would expect the datanodes to not need the ozone.metadata.dirs to be 
set, only to use it as a fallback
   
   Yes @neils-dev this existing part of the code is definitely not desirable. 
We should try to get to a point where `ozone.metadata.dirs` is truly optional, 
but that is tangental to the change in this PR, since this config is not being 
removed here. Let's continue a discussion of how to reach that point in a 
separate Jira. Probably related to HDDS-2025.
   
   For this change the `CRLDBDefinition` LGTM since it only used on the 
datanode, but I have a question on the `SecurityConfig`. Say there's a cluster 
with `ozone.metadata.dir` defined in the datanode config only, and neither 
`ozone.metadata.dir` or `hdds.metadata.dir` defined on the SCM, so its 
certificate info is written to `hdds.datanode.dir`.
   
   1. What will happen to the cluster when it is restarted with this change? 
The user would be prompted to set one of the existing configs, but what if they 
set it to a different directory than SCM was previously using, so SCM has no 
cert information? @fapifta do you know how the primordial SCM will handle this?
   
   2. Even if the above causes a problem, is it worth handling this somehow? It 
seems like a corner case and not worth leaving the `hdds.datanode.dir` config 
check in just for that.  I just want to make sure it is recoverable somehow if 
it does happen.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


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

Reply via email to