[GNC] Using Gnucash to access file over flaky network

[Stephen M. Butler]

On 11/24/21 07:54, Chris Green wrote:
> On Wed, Nov 24, 2021 at 11:24:21AM -0400, Chris Mitchell wrote:
>> Hi all,
>>
>> For reasons that are complicated but not especially interesting, I
>> would like to run Gnucash on one machine, with the data file located on
>> a remote machine — with the added challenge that the network access
>> available to the Gnucash "client" machine is a terrible cellular data
>> connection that sometimes drops without warning.
>>
>> I have daily backups, so I don't need strong guarantees against data
>> loss, but if it's possible, I'd like to set things up so there's
>> reasonable resilience against a network dropout corrupting the remote
>> Gnucash data. I'm not the only user with access to the data, so (given
>> that multi-user is still a long way off), I need file locking to work.
>> Having to manually delete an orphaned lock file after a network dropout
>> is acceptable.
>>
>> I assume that any of the database server backends would include this
>> kind of resilience "out of the box", and I'm not entirely unwilling to
>> try my hand at setting that up, but I am by no means a qualified
>> database administrator. If I can get sufficient resilience by easier
>> means, I'd prefer to stay away from the whole database server thing.
>>
>> What about Sqlite over sshfs?  I realize Sqlite is not designed for
>> access to a database residing on a different machine, but my inexpert
>> impression is that its "atomic commit" implementation should protect
>> against sudden disconnection between the program and the storage medium
>> just like it protects against sudden power loss. (IE the transaction
>> that's in the midst of being written will be lost, but the database
>> should be fine.)
>>
>> Can anyone confirm whether it's reasonable to expect that Gnucash with
>> Sqlite backend over sshfs would have working locking and decent
>> resilience against data corruption in this scenario? Or point out any
>> obvious "gotcha" I'm missing?
>>
> I think a better approach might be to simply copy the database to the
> machine you're working on when you start GnuCash and then copy the
> database back again afterwards.
>
Go so far as to create a script that would check for your own lock file 
before making the copy.  If lock file not present, then create it and 
copy the database.  Lock file could be as simple as renaming file with 
an extension of "xxx.IhaveIT".

Then the reverse script would remove the lock after copying the updated 
file back to the original name.

That way anybody else trying to update the file would know it was 
already "checked out".

-- 
Stephen M Butler, PMP, PSM
Stephen.M.Butler51 at gmail.com
kg7je at arrl.net
253-350-0166
-------------------------------------------
GnuPG Fingerprint:  8A25 9726 D439 758D D846 E5D4 282A 5477 0385 81D8