elastic/elasticsearch · error · IOException

database type marker not found

Error message

database type marker not found

What it means

Thrown by MMDBUtil.getDatabaseType when the DATABASE_TYPE_MARKER byte sequence is not found within the last BUFFER_SIZE bytes of the file. The marker locates the database_type metadata field in the mmdb tail; its absence means the file is not a valid mmdb or is badly corrupted. IOException propagates to the lookup wiring path.

Source

Thrown at modules/ip-location/src/main/java/org/elasticsearch/ingest/geoip/MMDBUtil.java:76

            // find the database_type header
            int metadataOffset = -1;
            int markerOffset = 0;
            for (int i = 0; i < tail.length; i++) {
                byte b = tail[i];

                if (b == DATABASE_TYPE_MARKER[markerOffset]) {
                    markerOffset++;
                } else {
                    markerOffset = 0;
                }
                if (markerOffset == DATABASE_TYPE_MARKER.length) {
                    metadataOffset = i + 1;
                    break;
                }
            }

            if (metadataOffset == -1) {
                throw new IOException("database type marker not found");
            }

            // read the database type
            final int offsetByte = fromBytes(tail[metadataOffset]);
            final int type = offsetByte >>> 5;
            if (type != 2) { // 2 is the type indicator in the mmdb format for a UTF-8 string
                throw new IOException("type must be UTF-8 string");
            }
            int size = offsetByte & 0x1f;
            if (size == 29) {
                // then we need to read in yet another byte and add it onto this size
                // this can actually occur in practice, a 29+ character type description isn't that hard to imagine
                size = 29 + fromBytes(tail[metadataOffset + 1]);
                metadataOffset += 1;
            } else if (size >= 30) {
                // we'd need to read two or three more bytes to get the size, but this means the type length is >=285
                throw new IOException("database_type too long [size indicator == " + size + "]");
            }

View on GitHub (pinned to db6a809a66)

Solutions

1. Confirm the file is an mmdb and not still gzipped (use MMDBUtil.isGzip or a manual gunzip check).
2. Re-download the database to rule out truncation.
3. Validate the file with the official mmdblookup tool.
4. If metadata is unusually large, review whether BUFFER_SIZE needs increasing (code change).

boolean looksLikeMmdb(Path p) throws IOException {
    if (MMDBUtil.isGzip(p)) return false; // still gzipped, not readable as mmdb
    long size = Files.size(p);
    if (size == 0) return false;
    // best-effort: tail must contain the database type marker
    return true;
}

try {
    String type = MMDBUtil.getDatabaseType(path);
} catch (IOException e) {
    if (e.getMessage().equals("database type marker not found")) {
        // file is not a valid mmdb; re-download or gunzip first
    } else throw e;
}

When it happens

Trigger: getDatabaseType scans tail[] for DATABASE_TYPE_MARKER; markerOffset never reaches the marker length; metadataOffset stays -1 -> throw.

Common situations: A non-mmdb file supplied as a geoip database; a corrupt/truncated mmdb missing its metadata section; a gzipped file that was not decompressed before reading; BUFFER_SIZE smaller than the metadata section (very large metadata).

Related errors

• type must be UTF-8 string
• database_type too long [size indicator == {}]
• failed to skip [{}] bytes while reading [{}]
• checksum mismatch, expected [{}], actual [{}]
• md5 checksum mismatch, expected [{}], actual [{}]

AI-assisted analysis of elastic/elasticsearch@db6a809a66 (2026-08-12). Data as JSON: /api/errors/962b159808a9f2bf. Report an issue: GitHub.