[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: subversion fud...and rcs old timers..

From: Max Bowsher <maxb_at_ukf.net>
Date: 2004-07-02 21:17:41 CEST

Brian Mathis wrote:
> Max Bowsher wrote:
>
>> Lindsay, Phil wrote:
>>
>>
>>> We are in the midst of conversion from clear case to subversion.
>>> Apparently someone in-house built a version of subversion on fedora 1
>>> (probably incorrectly?) and had a issue with a wedged/locked db file. This
>>> raised concerned form rcs folk about the security of source code
>>> should a db file become corrupted, etc...
>>>
>>>
>>
>> A wedged db hasn't lost any data.
>>
>>
>>
>>> Folk read Ben's fud note, but there is still some fear, if a db file is
>>> partially damaged is it recoverable sans backup?
>>>
>>>
>>
>> Depends on the extent and type of damage.
>>
>>
>>
>>> Have there been any
>>> issues between subversion versions and db structure?
>>>
>>>
>>
>> I don't understand the question.
>>
>>
>>
>>> Looking for any
>>> info that can help dispel the fear in some old time rcs folk (yes,
>>> they used to hack rcs files and are looking for analog to vi
>>> + rcs files).
>>>
>>>
>>
>> Well, If it really comes to that, the equivalent to vi is a scripting
>> language with berkeley DB bindings.
>>
>> But, seriously, all you have to do is implement a sensible backup regime.
>>
>> You can even use post-commit hooks to save incremental dumps of new
>> revisions as they are committed, until the next actual backup.
>>
>> Max.
>>
>>
>>
>
> You must admit, it's a little hard to sell something when it comes with
> the stipulation "You really, REALLY need to have good backups because
> something can go wrong, and probably will go wrong".

That wasn't what I intended to imply. Let me clarify:

All have to do, *to allay the last lingering doubts in your team*, is implement a sensible backup regime.

> Personally I think the DB backend is much more touchy than even the
> backend files for other things like postgresql - i think mostly due to
> the multiple access methods. If you stick to one and only one access
> method, either through svnserve or apache, the problems are reduced.
>
> Having said that, 1.1 will give you the option to use a file system
> backend, which I'm betting will be much more robust.

We've had _occasional_ reports of corruptions, its true. Only a fraction of those have been truly severe. Also, bear in mind that
you are likely to find out about a corruption, whereas with cvs, it's likely to go undetected.

Also, bear in mind that fsfs hasn't been released yet, so the vast majority of successful subversion users are using the bdb
backend.

Max.

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Fri Jul 2 21:23:08 2004

This is an archived mail posted to the Subversion Users mailing list.