Write-through proxy: easy slave-server de-synchronization

[Sergey Proskurnya <al...@gmail.com>]

Hello all,

I've examined write-through proxy feature and
found that the state of Slave server could be easily
de-synchronized (against the repo state of Master-server).

I've figure out that if someone commits to Master while
there is no connection between Slave and Master (ISP failure?),
then sample post-commit hook fails to synchronize repo state.
And after such event, all subsequent attempts will fail because
incremental "dump load" will fail also (or Slave will just loose one
revision?)


Do I have right understanding of described things?
If so, users must be warned about usage of sample post-commit hook
in real life environments with such distributed configuration.

Of cause, I'm going to develop home-brew shell script,
which lacks certain drawback, but I'd prefer to have
Master->Slave synchronization built-in in SVN binary or
implemented as standalone synchronization server ;-)

Thanks,
Serge.

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org

Re: Write-through proxy: easy slave-server de-synchronization

[Dave Lawrence <dl...@ad-holdings.co.uk>]

Mark Phippard wrote:
> On Tue, Jul 15, 2008 at 11:39 AM, Sergey Proskurnya <al...@gmail.com> wrote:
> 
>> I've examined write-through proxy feature and
>> found that the state of Slave server could be easily
>> de-synchronized (against the repo state of Master-server).
>>
>> I've figure out that if someone commits to Master while
>> there is no connection between Slave and Master (ISP failure?),
>> then sample post-commit hook fails to synchronize repo state.
>> And after such event, all subsequent attempts will fail because
>> incremental "dump load" will fail also (or Slave will just loose one
>> revision?)
> 
> I'd suggest you use svnsync to maintain the slave.  It will be smart
> enough to send all revisions it is missing.  I think the examples were
> written while svnsync was still in development.
> 

I just looked in the book and that appears to be more up to date on this 
issue than the release notes.  The release notes do kind of suggest that 
dump/load is the preferred method ("svnsync can probably do the same thing")



---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org

Re: Write-through proxy: easy slave-server de-synchronization

[Dave Lawrence <dl...@ad-holdings.co.uk>]

Mark Phippard wrote:
> On Tue, Jul 15, 2008 at 11:39 AM, Sergey Proskurnya <al...@gmail.com> wrote:
> 
>> I've examined write-through proxy feature and
>> found that the state of Slave server could be easily
>> de-synchronized (against the repo state of Master-server).
>>
>> I've figure out that if someone commits to Master while
>> there is no connection between Slave and Master (ISP failure?),
>> then sample post-commit hook fails to synchronize repo state.
>> And after such event, all subsequent attempts will fail because
>> incremental "dump load" will fail also (or Slave will just loose one
>> revision?)
> 
> I'd suggest you use svnsync to maintain the slave.  It will be smart
> enough to send all revisions it is missing.  I think the examples were
> written while svnsync was still in development.
> 

I just looked in the book and that appears to be more up to date on this 
issue than the release notes.  The release notes do kind of suggest that 
dump/load is the preferred method ("svnsync can probably do the same thing")



---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org

Re: Write-through proxy: easy slave-server de-synchronization

[Mark Phippard <ma...@gmail.com>]

On Tue, Jul 15, 2008 at 11:39 AM, Sergey Proskurnya <al...@gmail.com> wrote:

> I've examined write-through proxy feature and
> found that the state of Slave server could be easily
> de-synchronized (against the repo state of Master-server).
>
> I've figure out that if someone commits to Master while
> there is no connection between Slave and Master (ISP failure?),
> then sample post-commit hook fails to synchronize repo state.
> And after such event, all subsequent attempts will fail because
> incremental "dump load" will fail also (or Slave will just loose one
> revision?)

I'd suggest you use svnsync to maintain the slave.  It will be smart
enough to send all revisions it is missing.  I think the examples were
written while svnsync was still in development.

-- 
Thanks

Mark Phippard
http://markphip.blogspot.com/

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org