[
https://issues.apache.org/jira/browse/HDDS-15025?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18082508#comment-18082508
]
Ivan Andika edited comment on HDDS-15025 at 5/21/26 8:05 AM:
-------------------------------------------------------------
The main issue is that the admin updates a configuration by creating a new
configuration directory (e.g. /etc/ozone-om-new) and uses symlink to point
/etc/ozone-om from old directory (/etc/ozone-om-old) to new directory
(/etc/ozone-om-new). However, when existing OM node reload the configuration,
it still reloads the old configuration /etc/ozone-om-old even if the symlink of
/etc/ozone-om is already pointed to /etc/ozone-om-new. It's a JVM classloader
implementation will resolve the symlink location to the real location, any
reload will reload the resource from the real location.
So the solution is to not use symlink or change the SOP to create a backup file
first and change the configuration file in-place.
was (Author: JIRAUSER298977):
The main issue is that the admin updates a configuration by creating a new
configuration directory (e.g. /etc/ozone-om-new) and uses symlink to point
/etc/ozone-om from old directory (/etc/ozone-om-old) to new directory
(/etc/ozone-om-new). However, when existing OM node reload the configuration,
it still reloads the old configuration /etc/ozone-om-old even if the symlink of
/etc/ozone-om is already pointed to /etc/ozone-om-new.
So the solution is to not use symlink or change the SOP to create a backup file
first and change the configuration file in-place.
> Bootstrapped OM node is different than the Ratis configuration
> --------------------------------------------------------------
>
> Key: HDDS-15025
> URL: https://issues.apache.org/jira/browse/HDDS-15025
> Project: Apache Ozone
> Issue Type: Bug
> Reporter: Ivan Andika
> Assignee: Ivan Andika
> Priority: Minor
>
> We encountered issues where the bootstrapped OM returned in getServiceList
> ("ozone admin om roles") is different than the one in Ratis configuration.
> Reproduction (let's use newOmNode as the OM node ID of the bootstrapped OM)
> # Start all existing OM with wrong newOmNode node address configuration
> # Copy the wrong configuration to another file (e.g. wrong-configuration.xml)
> # Correct the new ozone-site.xml
> # Bootstrap the new OM newOmNode
> # The result is that the Ratis configuration has the correct address, but
> the OM node address is wrong
> I suspect the issue is that the OM is still using the old wrong configuration
> (wrong-configuration.xml) since OM does not set the
> OzoneManager.configuration when reloadConfiguration in getOmConfiguration.
> This creates a mismatch between Ratis and OM states.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
