This task tracks moving production to dumps-nfs.wikimedia.org and its /mnt/nfs/dumps mountpoint.
Unfortunately some code in production hardcodes /mnt/nfs/dumps-clouddumps1001.wikimedia.org and /mnt/nfs/dumps-clouddumps1002.wikimedia.org. Which we'll need to migrate to their canonical /mnt/data/xmldatadumps/public location instead. In production that is a symlink which points to the active clouddumps host. In the final state it will instead point to /mnt/nfs/dumps which is the dumps-nfs.wikimedia.org mount location. Upon failover umount and re-mount of said location is taken care automatically.
Hardcoded path locations
- https://gerrit.wikimedia.org/g/operations/cookbooks/+/9b41f5767a9c1992a24ef53ae05f743d78ed9ec6/cookbooks/sre/wdqs/data-reload.py
- https://gitlab.wikimedia.org/repos/wikidata-platform/triplestores/-/tree/ae4ebbda8f90633e44e84f56b923c405669c0015/qlever/README.md
- https://gitlab.wikimedia.org/repos/wikidata-platform/triplestores/-/tree/ae4ebbda8f90633e44e84f56b923c405669c0015/virtuoso/bin/split.sh
- https://gitlab.wikimedia.org/repos/data-engineering/airflow-dags/-/tree/68cff9f54cf1f553394404fba888b9fe4884d194/platform_eng/dags/detect_html_tables_dag.py
- https://gitlab.wikimedia.org/repos/data-engineering/airflow-dags/-/tree/68cff9f54cf1f553394404fba888b9fe4884d194/tests/platform_eng/fixtures/bash_operator/platform_eng_dags_detect_html_tables_dag.py-detect_html_tables-move_dumps_to_hdfs.expected
Production hosts mounting dumps-nfs as of Jul 15th 2026
- an-launcher1003
- stat1008
- stat1009
- stat1010
- stat1011
- wcqs2001
- wdqs1024
- wdqs1029
- wdqs1030
- wdqs1031
- wdqs1032
- wdqs2010