After upgrading VMware Aria Operations for Logs from version 8.16 to 8.18.5, the User Interface (UI) is inaccessible. The Cassandra service fails to start, and logs show the following error pattern:
ERROR [SSTableBatchOpen:1] SSTableReader.java:562 - Cannot read sstable /storage/core/loginsight/cidata/cassandra/data/system/...; file system error, skipping tableorg.apache.cassandra.io.FSReadError: java.nio.charset.MalformedInputException: Input length = 1
Aria Operations for logs: 8.18.x
This issue is caused by filesystem corruption on the virtual appliance nodes, leading to corrupted Cassandra SSTable structures or migration metadata. Corruption typically occurs due to:
Resolution Steps:
Ensure you have a valid backup or offline snapshots of all cluster nodes before proceeding with these steps.
How to take a Snapshot of Operations for Logs
Identify the corrupted directories mentioned in the Cassandra logs. Move these directories to a temporary location outside of the Cassandra data path to allow the service to initialize while skipping the unreadable tables.
If the /storage/core/loginsight/cidata/cassandra/migrations directory is corrupted, it must be restored by copying the directory from a healthy peer node in the cluster.
On nodes where corruption is widespread and preventing service startup, clear the local log store using the following steps:
This command will make changes to your system. Review it carefully before running.
# Stop the Log Insight serviceservice loginsight stop
# Navigate to the log store directorycd /storage/core/loginsight/cidata/store
# Remove all contents within the store directoryrm -rf *
# Restart the service to re-initialize the storage pathservice loginsight start