Cloud authentication isn't renegotiated on long backups
| Affects | Status | Importance | Assigned to | Milestone | |
|---|---|---|---|---|---|
| Duplicity |
New
|
Undecided
|
Unassigned | ||
Bug Description
Version 0.7.10
Python 2.7.9
Debian 8.6 (jessie)
Tested backends: Backblaze, Hubic
I have a couple of large backups of around 300GB and 500GB that I am trying to create on cloud storage. I have tested with Backblaze B2 and Hubic as both seem to offer good value for money and as they work with duplicity. Both have token based authentication that times out after a period - on Backblaze it seems to be 12 hours and on Hubic it seems to be 48. Neither backend renegotiates once it starts failing. I would imagine that this applies to other cloud services.
A common issue is that negotiation occurs in the init function and isn't tried again once the cloud service rejects the connection when the authentication period times out. A fix for this should be that the negotiation function is separated out and then can be called when the service rejects the connection, usually with a 403 error. I have started work on this for Backblaze.
However, it also seems, certainly with the Backblaze backend, that resuming a failed backup doesn't initialise the connection and that the authentication token isn't renegotiated so it may also be the same with other services.
| description: | updated |
