Skip to content

[hf] process destruction requests in sequence to avoid deadlock

aguestuser requested to merge hf-process-destruction-requests-in-sequence-to-avoid-deadlock into main
  • BUG: processing mature destruction requests always yields following error: Error processing destruction jobs: SequelizeConnectionAcquireTimeoutError: Operation timeout
  • POTENTIAL CAUSE: we launch several deletion jobs in parallel, all of which seek to take locks on the phone numbers table, potentially creating deadlock over the tx lock
  • ATTEMPTED FIX: try performing each destroy call in sequence to avoid contention over the lock. (sorta expensive/cludgy, but should work for a job that we don't run very often!)

Merge request reports