Re: [Duplicity-talk] Backend for Amazon S3

[Eric Evans]

[ Kenneth Loafman ]
> I've been looking into how to do that for all the backends, however,
> infinite retries is not the option.  In this case a simple change from
> run_command() to run_command_persist() allows retries.  I'll look into a
> model we can use to standardize error handling across all the backends.

In the meantime, would it be helpful if I submitted a new patch that
traps exceptions and tries again once? Since boto already retries once on
it's own, that'd make 4 tries total, one more than run_command_persist().

> BTW, the goal is for duplicity to run in the background, unattended, via
> scheduling.  Backups should be regular and scheduled.  You can't do that
> if you rely on humans to initiate them.

Yeah, I run it from cron.

> Charles Knowlton wrote:
> > Maybe there could be an option for how many times you want it to retry
> > that way the user
> > has more control over how many times it retries instead of it be
> > indefinitely
> > 
> > Charles
> > 
> > On Jul 8, 2007, at 1:29 AM, address@hidden wrote:
> > 
> >> Eric Evans wrote:
> >>> Yes, boto will retry once before raising an exception. The only
> >>> situations I've personally encountered where that wasn't enough
> >>> were outright outages on Amazon's end.
> >>> By contrast, the bitbucket backend will retry every 10 seconds
> >>> indefinately, (it seems to be the only backend that does that). I'm not
> >>> sure how others feel about that, but it might be rather annoying if
> >>> duplicity were being run non-interactively and the error condition
> >>> persisted.
> >>>
> >>
> >> I submitted the patch that makes the current S3 backend retry
> >> indefinitely because a single retry consistently failed during my
> >> backups.  I agree that the current behavior may be annoying, and can
> >> be improved, but a single retry isn't sufficient.

-- 
Eric Evans
address@hidden

signature.asc
Description: Digital signature