jackrabbit-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Raphaël Franchet <raph...@anyware-tech.com>
Subject Re: Repository inconsistent with xmlpersistentmanager
Date Mon, 22 May 2006 08:03:57 GMT
Hello,

For information, the source of the problem with the XMLPersistantManager 
has been found:
As I told you before, the repository is huge (20 GB) and it reached the 
limit of the subdirectory per directory on linux (32k directories) : so 
sometimes it worked (if the hash key was already existing) some times 
not (if the hash key was new and needed a new directory)
Now we've found the problem we can imagine solutions.

My question is : why can't we find any error message in any log file ?
If a directory cannot be created I thought we would find an error !

Rgd,

Raphaël FRANCHET
Chef de projet
Solutions & Technologies
ANYWARE TECHNOLOGIES
Tel: +33 (0)5 61 00 73 47
Fax : +33 (0)5 61 00 51 46
www.anyware-tech.com


Nicolas a écrit :
> Hey Cedric,
> 
> I think the best in your case would be to export/import your data from 
> your old repository to a new one with a new persistence manager.
> 
> The BerkeleyDBPM has slightly better performance, but the license may be 
> a problem.
> The DerbyPM sounds like a good open-source enterprise solution.
> 
> Regards,
> 
> Nicolas,
> 
> 
> On May 20, 2006, at 3:05 AM, Cédric Damioli wrote:
> 
>> Stefan Guggisberg a écrit :
>>> hi raphaël
>>>
>>> On 5/19/06, Raphaël Franchet <raphael@anyware-tech.com> wrote:
>>>> Hello,
>>>>
>>>> I'm using JR 1.0 with the xmlPersistanceManager.
>>>
>>> the XMLPersistenceManager is not recommended to be used
>>> in a production environment. its major flaw , apart from being dead 
>>> slow ;),
>>> is that it is 'non-transactional', i.e. if your jvm process suffers a
>>> sudden death
>>> (power outage, kill, etc) you'll very likely end up with an
>>> inconsistent repository
>>> state (like the one you described).
>>>
>>> by using jackrabbit's default configuration (i.e. 
>>> DerbyPersistenceManager)
>>> you avoid such issues.
>>>
>>>
>> Hi Marcel,
>>
>> We know that "now" the XMLPM is not the recommanded one, but as early 
>> adopters of Jackrabbit, more than one year ago, there was only XMLPM 
>> and ObjectPM :-)
>>
>> We now have an inconsistent repository. What would you suggest ? 
>> Migration to a new PM (if it is possible) ? Debugging of our current 
>> repository ?
>>
>> BTW, what would you recommend as a new PM : DerbyPM or BerkeleyDBPM ?
>>
>> Regards,
>> Cédric
>>
>>>
>>>> My application has many nodes (about 20 GB of data) and I've got a huge
>>>> problem : my repository is going inconsistent.
>>>>
>>>> Using the API normally, I obtain some time the following problem :
>>>> On the file system, node 'A' references a child node 'B' (in his xml
>>>> persistent state) but node 'B' does not exist (there is no 
>>>> corresponding
>>>> xml file).
>>>> The API is telling me that node A has no child, but if I try to add 
>>>> B it
>>>> fails (telling that it can't find the already referenced B).
>>>>
>>>> The problem appeared once yesterday, so we restored a daily backup 
>>>> (that
>>>> is clean on this node) but after a day of use the same problem is 
>>>> coming
>>>> up (several times on the same repository)
>>>>
>>>> The only particularity I've noticed on the problematic node, is that it
>>>> is removed and re-added in the same session, quite often.
>>>> The application is used by many people at a time that may be accessing
>>>> the same nodes.
>>>>
>>>> Any experience or idea ?
>>>> (considering I don't believe this is an astral conjunction)
>>>>
>>>> Regards,
>>>>
>>>> -- 
>>>> Raphaël FRANCHET
>>>> Chef de projet
>>>> Solutions & Technologies
>>>> ANYWARE TECHNOLOGIES
>>>> Tel: +33 (0)5 61 00 73 47
>>>> Fax : +33 (0)5 61 00 51 46
>>>> www.anyware-tech.com
>>>>
>>
>>
>> --Cédric Damioli
>> Chef de projets systèmes d'informations
>> Solutions CMS
>> ANYWARE TECHNOLOGIES
>> Tel : +33 (0)5 61 00 52 90
>> Fax : +33 (0)5 61 00 51 46
>> http://www.anyware-tech.com
> 

Mime
View raw message