apache/flink · critical · FlinkException

Could not cancel job {}.

Error message

Could not cancel job {}.

What it means

Thrown from AbstractColumnReader.readDataBuffer when slicing `length` bytes from the page's data stream fails — i.e. the page has fewer remaining bytes than the fixed-width value being read requires. The message states exactly how many bytes were requested. This is a page-body truncation invariant violation.

Source

Thrown at flink-clients/src/main/java/org/apache/flink/client/cli/CliFrontend.java:670

                                + targetDirectory
                                + '.');
            }

            runClusterAction(
                    activeCommandLine,
                    commandLine,
                    (clusterClient, effectiveConfiguration) -> {
                        final String savepointPath;
                        try {
                            savepointPath =
                                    clusterClient
                                            .cancelWithSavepoint(jobId, targetDirectory, formatType)
                                            .get(
                                                    getClientTimeout(effectiveConfiguration)
                                                            .toMillis(),
                                                    TimeUnit.MILLISECONDS);
                        } catch (Exception e) {
                            throw new FlinkException("Could not cancel job " + jobId + '.', e);
                        }
                        logAndSysout(
                                "Cancelled job "
                                        + jobId
                                        + ". Savepoint stored in "
                                        + savepointPath
                                        + '.');
                    });
        } else {
            final JobID jobId;

            if (cleanedArgs.length > 0) {
                jobId = parseJobId(cleanedArgs[0]);
            } else {
                throw new CliArgsException("Missing JobID. Specify a JobID to cancel a job.");
            }

            logAndSysout("Cancelling job " + jobId + '.');

View on GitHub (pinned to 2f3c205e92)

Solutions

1. Compare the requested length in the message with the page size — a fixed-width read exceeding remaining bytes proves page truncation
2. Re-obtain the file and verify its size/CRC; retry read on a clean copy
3. Dump the column chunk with `parquet-tools dump -c <column>` to confirm the page layout in a reference reader
4. If reproducible, the writer produced invalid pages — capture the file and report to the writer's project

// Before reading fixed-length types, confirm page byte counts from metadata
ColumnChunkMetaData cc = ...;
int typeLen = cc.getType() == PrimitiveType.PrimitiveTypeName.FIXED_LEN_BYTE_ARRAY
        ? cc.getTypeLength() : cc.getType().byteSize();
// page sizes live in the page index / column index when present; absent them,
// rely on footer-consistency checks: encodings + total size must be coherent

try {
    vectorizedReader.readToVector(...);
} catch (ParquetDecodingException e) {
    if (String.valueOf(e.getMessage()).startsWith("Failed to read")) {
        quarantine(file); // deterministic truncation
    } else throw e;
}

When it happens

Trigger: Reading FIXED_LEN_BYTE_ARRAY (e.g. DECIMAL, INT96 timestamps) or fixed-width values where the remaining page bytes < value length; page metadata count/values inconsistent with actual page size; corrupted page bytes.

Common situations: Truncated files in object storage; writers with bugs emitting short pages for fixed-length types; decimal(38,x) or INT96 timestamp columns in malformed files.

Related errors

• Error while waiting for job to be initialized
• Could not build the program from JAR file: {}
• Missing JobID. Specify a JobID to cancel a job.
• Failed to dispose the savepoint '{}'.
• Failed to read from input stream

AI-assisted analysis of apache/flink@2f3c205e92 (2026-08-14). Data as JSON: /api/errors/08ef92294b437553. Report an issue: GitHub.