1591881 : Misuse vom Maxwell Cluster

Created: 2026-04-14T06:24:10Z - current status: new

Here is the anonymized and summarized report of the issue:

Summary of the Issue

The Maxwell cluster is generating repeated, automated requests to the FATER system (a document delivery service for DESY publications). These requests appear in rapid succession (every ~5 seconds), but each query takes approximately 3 minutes to complete. This high frequency of automated queries is overloading both the FATER server and its network infrastructure.

Key Observations: - The requests originate from an IP address (131.169.178.156) associated with the Maxwell cluster. - Each request includes a DOI (doi/10.3204/PUBDB-2020-00969), suggesting an attempt to fetch publication metadata or full texts. - The system logs indicate broken pipe errors (Apache2::RequestIO::rflush), likely due to timeouts or resource exhaustion. - FATER is not designed for bulk automated downloads but for manual, individual queries.

Suggested Solutions/Next Steps

  1. Identify the Source of the Requests
  2. Check which user account or job script on the Maxwell cluster is generating these queries.
  3. Review recent job submissions (e.g., via sacct or squeue) to trace the origin of the requests.

  4. Example command to list recent jobs: bash sacct --format=JobID,User,JobName,Start,End,State --starttime=2026-04-14

  5. Implement Rate Limiting or Delays

  6. If the requests are part of a legitimate workflow (e.g., a script fetching publication data), modify the script to:
    • Add delays between queries (e.g., sleep 300 for 5-minute intervals).
    • Use batch processing with controlled concurrency (e.g., xargs -P 1 for sequential requests).

  7. Example for a Python script: python import time time.sleep(300) # Wait 5 minutes between requests

  8. Use Alternative Data Sources

  9. For bulk metadata/full-text downloads, consider:

    • DESY’s internal publication database (e.g., PUBDB) instead of FATER.
    • APIs (if available) for programmatic access to DESY publications.
    • Local mirrors of frequently accessed datasets.

  10. Temporary Blocking (if Necessary)

  11. As a last resort, temporarily block the Maxwell cluster’s IP (131.169.178.156) from accessing FATER until the issue is resolved.

  12. Example firewall rule (adjust for DESY’s infrastructure): bash iptables -A INPUT -s 131.169.178.156 -p tcp --dport 80 -j DROP

  13. User Communication

  14. Notify the user(s) responsible for the requests to:
    • Acknowledge the impact on FATER.
    • Provide guidance on compliant alternatives (e.g., using PUBDB or APIs).

References