jackrabbit-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Noah Vihinen <nvihi...@maven.net>
Subject Re: Removing Index
Date Wed, 06 Jun 2007 03:08:07 GMT
Opened https://issues.apache.org/jira/browse/JCR-964 as this seems to  
be a blocker for us, until we can figure out how to rebuild an index  
from information in our DB.

On Jun 5, 2007, at 7:21 AM, Noah Vihinen wrote:

> OK, that makes sense.  However my main question is how do I recover  
> from a corrupt search index?  The automatic repair that seems to  
> kick in on a restart notifies me that the index cannot be  
> repaired.  It seems like a DB backup is not enough, and I also have  
> to back up my file-based search indexes.  This seems bad, as it  
> will be nearly impossible to take these two backup snapshots of two  
> persistent stores in the same state.
>
> On Jun 4, 2007, at 5:05 AM, Marcel Reutegger wrote:
>
>> Hi Noah,
>>
>> Noah Vihinen wrote:
>>> We've been running into issues with our jackrabbit index on  
>>> forced shutdowns.  On restarts, we end up getting errors like the  
>>> following:
>>> Error indexing root node:
>>> java.io.IOException: Error indexing root node:  
>>> b5d9c4a1-07d5-471e-9fc9-e9a2fbd6ae48
>>>         at  
>>> org.apache.jackrabbit.core.query.lucene.MultiIndex.<init> 
>>> (MultiIndex.java:313)         at  
>>> org.apache.jackrabbit.core.query.lucene.SearchIndex.doInit 
>>> (SearchIndex.java:295)         at  
>>> org.apache.jackrabbit.core.query.AbstractQueryHandler.init 
>>> (AbstractQueryHandler.java:44)
>> [...]
>>> We attempted to get around this issue by simply deleted the index  
>>> of the repository (on the particular machine in a cluster) and  
>>> restarted with no luck.  In the end, the only thing that seems to  
>>> work is to copy the index directory from one of the other  
>>> functional machines in our cluster, followed by a restart.
>>> This seems like a really ugly way of getting around this issue.   
>>> Is there something obvious we're doing wrong.  Shouldn't it be  
>>> possible to add a machine to a cluster with no pre-existing  
>>> index?  What are we missing?
>>
>> My guess is, that this happens because content is changed while  
>> the new cluster node indexes (traverses) the workspace.
>>
>> Might be related to: http://issues.apache.org/jira/browse/JCR-931
>>
>> regards
>>  marcel
>


Mime
View raw message