apache/cassandra · warning

Failed to delete {}

Error message

Failed to delete {}

What it means

When saving a cache (key cache / row cache), AutoSavingCache.deleteOldCacheFiles cleans up stale files in the saved_caches directory that match this cache's name or type. If File.tryDelete() fails to unlink such a file, this warning is logged and the cleaner continues. Stale file remains on disk but cache save/load still proceeds.

Source

Thrown at src/java/org/apache/cassandra/cache/AutoSavingCache.java:453

        private void deleteOldCacheFiles()
        {
            File savedCachesDir = new File(DatabaseDescriptor.getSavedCachesLocation());
            assert savedCachesDir.exists() && savedCachesDir.isDirectory();
            File[] files = savedCachesDir.tryList();
            if (files != null)
            {
                String cacheNameFormat = String.format("%s-%s.db", cacheType.toString(), CURRENT_VERSION);
                for (File file : files)
                {
                    if (!file.isFile())
                        continue; // someone's been messing with our directory.  naughty!

                    if (file.name().endsWith(cacheNameFormat)
                     || file.name().endsWith(cacheType.toString()))
                    {
                        if (!file.tryDelete())
                            logger.warn("Failed to delete {}", file.absolutePath());
                    }
                }
            }
            else
            {
                logger.warn("Could not list files in {}", savedCachesDir);
            }
        }

        public boolean isGlobal()
        {
            return false;
        }
    }

    /**
     * A base cache serializer that is used to serialize/deserialize a cache to/from disk.
     * <p>

View on GitHub (pinned to 88fd0f6a0e)

Solutions

1. Stop Cassandra and manually delete stale files in the saved_caches directory (they are disposable caches, safe to remove while the node is down).
2. Fix ownership/permissions: chown -R cassandra:cassandra /var/lib/cassandra/saved_caches.
3. Check disk space and mount status (df -h, mount) — a full or read-only filesystem makes unlink fail.
4. Ensure no external processes (backups, antivirus) hold those files open when Cassandra runs.

Example fix

// before (bash): fix leftover root-owned cache files
sudo rm -f /var/lib/cassandra/saved_caches/*
sudo chown -R cassandra:cassandra /var/lib/cassandra/saved_caches

// Pre-flight check on deployment (bash)
test -w /var/lib/cassandra/saved_caches || echo "saved_caches not writable by $(id -un)"
find /var/lib/cassandra/saved_caches ! -user cassandra -ls  # detect foreign-owned files

When it happens

Trigger: A file in saved_caches whose name ends with the cache's name format or cache type cannot be deleted — typically because the file is open by another process, or the Cassandra process lacks write/remove permission on the directory/file.

Common situations: Running Cassandra as a different user after a manual start (root-created files left behind); backup agents or file watchers holding files open; read-only or full filesystem; NFS-mounted saved_caches directories.

Related errors

• Could not list files in {}
• Unable check disk space in '%s'. Perhaps the Cassandra user
• ; unable to start server
• Unable to create directory " + dir
• Dictionary file %s is not readable.

AI-assisted analysis of apache/cassandra@88fd0f6a0e (2026-09-10). Data as JSON: /api/errors/b2dadb27fc21db56. Report an issue: GitHub.