In the NSX-T 2.3.x dashboard, you see disk usage alert for controller node similar to:
Disk usage is high for filesystem /dev/mapper/nsx-var+log
When you check the volume mount information on the controller, you see that /var/log/moot-server/Zookeeper_0 directory taking up most of the space.
You see that zookeeper.log, zookeeper-cli.log files does not get compressed.
The issue occurs as the log rotation does not work as expected.
This is a known issue with VMware NSX-T 2.3.0 and 2.3.1.
This issue is resolved in VMware NSX-T 2.4.1 and later.
To work around this issue in NSX-T 2.3.x, change the log file rotation parameters as mentioned below on the problematic Controller node:
Log in to Controller node as root user.
Edit the log4j.properties file located at:
/config/moot-server/localServiceManager/containers/Zookeeper/persistentStore/zkWorkspace
Update the following parameters:
zookeeper.log.maxfilesize=256MB
zookeeper.log.maxbackupindex=20
Set these values as per the requirement and free space on controller node.
Validate the changes:
# /var/log/moot-server/Zookeeper_0# grep "zookeeper.log.maxfilesize\|zookeeper.log.maxbackupindex" /config/moot-server/localServiceManager/containers/Zookeeper/persistentStore/zkWorkspace/log4j.properties
zookeeper.log.maxfilesize=128MB
zookeeper.log.maxbackupindex=5
Restart the controller to apply the changes.
Note: Changes on the log4j.properties will be persistent after the reboot of the controller only. Changes will not be reflecting if you do service restart like this /etc/init.d/moot-server restart
This will cause the control plane to malfunction as the moot service (which is a container for zookeeper) will stop. Due to this zookeeper may not start and control cluster will be impacted. So controller will not function as expected.