apache/cassandra · warning

Error loading key or row cache

Error message

Error loading key or row cache

What it means

Warning logged during CassandraDaemon.setup when asynchronous loading of the key and/or row cache from disk fails. JVMStabilityInspector.inspectThrowable decides whether the failure is fatal (e.g. disk corruption usually is); otherwise the node continues with cold caches.

Source

Thrown at src/java/org/apache/cassandra/service/CassandraDaemon.java:340

        }

        // We need to persist this as soon as possible after startup checks.
        // This should be the first write to SystemKeyspace (CASSANDRA-11742)
        SystemKeyspace.persistLocalMetadata();

        Thread.setDefaultUncaughtExceptionHandler(JVMStabilityInspector::uncaughtException);

        SystemKeyspaceMigrator41.migrate();
        setupVirtualKeyspaces();

        try
        {
            loadRowAndKeyCacheAsync().get();
        }
        catch (Throwable t)
        {
            JVMStabilityInspector.inspectThrowable(t);
            logger.warn("Error loading key or row cache", t);
        }

        if (!SKIP_GC_INSPECTOR)
        {
            try
            {
                GCInspector.register();
            }
            catch (Throwable t)
            {
                JVMStabilityInspector.inspectThrowable(t);
                logger.warn("Unable to start GCInspector (currently only supported on the Sun JVM)");
            }
        }

        // Replay any CommitLogSegments found on disk
        PaxosState.initializeTrackers();

View on GitHub (pinned to 88fd0f6a0e)

Solutions

1. Inspect the chained exception for the offending cache file and delete/clear the corrupt cache files from saved_caches_directory, then restart (node starts with cold cache)
2. Check filesystem permissions and disk health (dmesg, smartctl) for the cache directory volume
3. Verify cache settings in cassandra.yaml (key_cache_size_in_mb, row_cache_size_in_mb, saved_caches_directory) are consistent with the on-disk files
4. If corruption recurs, disable the row cache or resize caches to reduce exposure

// before start: check cache dir readable and files intact
ls -la /var/lib/cassandra/saved_caches
# and inspect any *.db cache files for truncation (size > 0, no partial temp files)

try {
    loadRowAndKeyCacheAsync().get();
} catch (Throwable t) {
    JVMStabilityInspector.inspectThrowable(t);
    logger.warn("Error loading key or row cache", t);
    // accept cold cache, or remove corrupt saved_caches files and restart
}

When it happens

Trigger: loadRowAndKeyCacheAsync().get() throws: cache files under saved_caches_directory are corrupt, unreadable (permissions), or the cache size config mismatches; disk I/O error while reading cache files; an InterruptedException during the blocking get().

Common situations: Corrupt or truncated saved_caches files after an unclean shutdown or disk-full event; wrong permissions on the saved_caches directory; moving cache files between nodes/versions; disk hardware failures.

Understand the failure class

Background: "failed to read file", EACCES, ENOENT and "could not read <path>" errors: when a program can't read a file from disk — this error's family across 49 libraries.

Related errors

• Error loading counter cache
• Cache schema version + expected + does not match current sch
• Cannot find configured row cache provider class ${DatabaseDe
• Corrupted key cache. Key length of %d is longer than maximum
• Corrupted key cache. Failed to deserialize key of key cache

AI-assisted analysis of apache/cassandra@88fd0f6a0e (2026-09-10). Data as JSON: /api/errors/539ee30f9ce8e933. Report an issue: GitHub.