Intermittent hang on large dblock download from OneDrive during compact phase (2.3.0.3)

My backup jobs to OneDrive intermittently hang during the post-backup compact phase (UI shows “Deleting unwanted files”). It’s not tied to a specific job — different jobs hit it at different times — which suggests a general OneDrive backend issue rather than a job-specific problem. The main upload completes, then the job stalls indefinitely and must be stopped manually. Profiling logs show it hangs on a download of a large dblock that starts but never completes.

Environment:

  • Duplicati 2.3.0.3 (LinuxServer.io Docker image, v2.3.0.3_stable)
  • Host: TrueNAS Scale, running in Docker
  • Destination: OneDrive
  • Update channel: Stable

Behavior:

  • Backup upload phase completes normally.
  • Job then stalls during the compact/cleanup phase.
  • Container CPU drops to ~0.08% — the process is idle/waiting, not CPU-bound.

sudo docker stats duplicati --no-stream
CONTAINER ID NAME CPU % MEM USAGE / LIMIT MEM % NET I/O BLOCK I/O PIDS
1546a8adfd21 duplicati 0.08% 929.2MiB / 125.7GiB 0.72% 6.07GB / 3.54GB 131kB / 0B 30

  • Profiling log shows: Backend event: Get - Started: duplicati-....dblock.zip.aes (249.830 MiB) with no corresponding “Get - Completed” and no error. Smaller dindex files (~16 KiB) in the same run complete in ~1s. The large download simply never returns.
  • Requires manual stop; on restart it goes straight back to the same phase.
  • Affects different jobs at different times, not one specific job/dataset.
  • I have to stop/abort the job or restart the docker container to cancel the hung job.

My thoughts: A large-file download from OneDrive intermittently stalls with no timeout, so the operation waits forever. Because it affects different jobs at different times, it appears to be a backend transfer issue rather than database or job-specific corruption.

Possibly related: Thread #22345 (“Completing previous backup”) describes a hang in the same post-backup phase — but that one is CPU-bound in SQLite (100% one core, slow AppendFilesFromPreviousSetAsync). Mine is the opposite: I/O-bound, ~0% CPU, waiting on a stalled OneDrive download. Same phase, different failure point — possibly the same “interrupted-backup handling is fragile in 2.3.x” area surfacing two different ways.

Questions:

  1. Is the stalled large-download a known OneDrive backend issue in 2.3.x?
  2. Would --read-write-timeout / --http-operation-timeout convert this hang into a retryable failure? Recommended values?
  3. Would --disable-synthetic-filelist help for jobs where partial-restore isn’t needed?

Duplicati error log.txt.zip (2.8 KB)

After 9 days, not one comment? I’ve had my backups hang 4 more times in the 9 days that it’s taken no one to respond. That’s not a solution I care to run. Onto to find a new alternative.