[Box Backup-dev] A new problem with diffing large files?

[Ben Summers]

On 14 Feb 2006, at 21:41, Gary wrote:

> Ben,
>
>> Does this still happen with the latest Win32 releases? Or just in
>> your original code?
>
> I cannot say, since there is no Win32 version yet with KeepAlive(),  
> yet (except for old Nick's
> release tree for 0.9).

What about the versions which Chris keeps on posting?

>
>> Has anyone actually tested the keep-alive stuff on UNIX platforms?
>
> I believe it will be the same thing. Well, according to the log:
>
> 1. KeepAlive()s do happen.
> 2. The error comes up after sending stream (NOT after diffing), and  
> happens upon Finished().
>
> Not sure what might be going on there, why would diffing work,  
> uploading work, and then Finished()
> fail? Makes no sense, there is constant network activity.

How do you know that this is the case? Have you checked the logs on  
the server?

You might see this error if the server ran out of space in the  
account, but you can only tell if you look at the logs on the server  
as well.

(Yes, error reporting from bbackupd and bbstored is not wonderful.)

>
> That is, unless sending a network-based KeepAlive() command does  
> not "bump" some part of the
> client/server procotol or SLL layers. Perhaps KeepAlive() should  
> actually transmit some data over
> the SLL link?

It does transmit data! There is no timeout in the Box Backup code,  
only in the underlying OS socket. Sending a few bytes is more than  
enough to convince the remote side that it's still being used.



On 14 Feb 2006, at 21:42, Gary wrote:
>> Has anyone actually tested the keep-alive stuff on UNIX platforms?
>
> At any rate, it looks like we need a long, detailed, hard look at  
> client/server procotol layer,
> SSL layer, and network layer timeout possibilities in the current  
> code...

The basic Protocol code has not changed.

Ben