[Box Backup] Connection fails during backup to remote server

[Matt Brown]

Hi Chris,

>> Well, I would envisage that once a snapshot has been started that  
>> it would
>> sync until its completion.
>
> Yeah, seems reasonable to me too. I can send you a patch to test if  
> you're
> interested.

Sure, I am happy to help testing as always :-)

>> In this instance the connection/socket is reset presumably by the 1  
>> & 1
>> firewall states which in turn cause the connection to be
>> severed/terminated ... obviously a couple of things can be re- 
>> arranged
>> to make this work - i.e iptables or alternative location.
>
> Seems like a good idea, but I agree that we should recover and try  
> again
> if the sync fails, even in snapshot mode (currently we don't).

Well on reflection, maybe people use the lazy mode more than  
snapshot ? maybe some feedback from other box users could confirm ? as  
this does provide a better recovery solution to files which are always  
changing, and presumably the connection would be automatically re- 
tried on the next lazy run ? in addition to any client retries  
currently coded for..

>> I am also guessing that once box incorporates mark states this will
>> require a snapshot type sync ?
>
> Sorry, what are "mark states"? Do you mean like a point-in-time  
> snapshot
> of all files on the filesystem at that time, to fix the deletion time
> problem? If so, then I think we could just do snapshots at irregular
> intervals, which is more or less exactly what we do now.

Sorry, yes my bad I was using the wording as per the Box Backup web  
page:

Panned future work

     * Mark state (marks the current state of the filesystem, to  
emulate a tape change and restoring from that tape later)

So far we have only had to restore the odd files here and there that  
people have deleted/corrupted.. however I am not sure how you would go  
about recovering so a whole directory from 2 days ago, is that  
possible currently ?

Please don't think I am knocking the project, its works very well and  
has been 100% successful in recovering data ! I wasnt sure if I had  
hit a bug, but would appear to be more of a feature request :-)

Matt