lucene-java-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Andrzej Bialecki>
Subject Re: Another index corruption problem
Date Wed, 26 Oct 2005 07:56:51 GMT
Bill Tschumy wrote:
> I hate to plead, but I really need to do my best to recover my  
> customer's data.  Does anyone have any pointers for how to manually  (or 
> programmatically) repair this corrupted index?
> On Oct 24, 2005, at 11:23 PM, Bill Tschumy wrote:
>> Many months ago I wrote this list about a corrupted index that one  of 
>> my customers had.  It was a mystery that was never really  solved.  
>> Well, it has happened again and the stack trace looks  almost 
>> identical.  Here is the exception:
>> /Users/samegan/Library/Preferences/ 
>> Parsnips/IndexData/_1d.fnm (No such file or directory)

This doesn't look good. Open the "segments" file with a hex editor and 
check it - is it just "_1d" there (apart from other unrecognizable stuff)?

You can try to manually "explode" the compound format into individual 
files, and then try to go from there - please look at 
org.apache.lucene.index.CompoundIndexReader how to read the compound 
format. In the best case, this could mean that you will have to recreate 
the .fnm file, in the worst case... well, you already know. :-/

>> This is a compound index, so I guess the .fnm file it is looking  for 
>> is internal to it or temporary in some way.  The customer  thinks the 
>> problem was caused by accidentally pasting an entire  document into 
>> the "title" field and saving.  I kind of doubt this  caused the 
>> problem, but you never know.  I treat the "title" and  the "body" 
>> identically for indexing.

No, this is very unlikely. What is more likely the cause is that the 
application (or machine) crashed during write or optimize.

>> The person is pretty panicked about his lost data.  Does anyone  have 
>> any hints as to how to edit the file to get it back  functioning 
>> again?  I've heard of people using hex editors for this.
>> After solving his immediate problem, I need to figure out why this  is 
>> happening.  I haven't followed this list for a couple of  months.  Has 
>> anything like this come up recently?  I am using  Lucene-1.4.3.


Best regards,
Andrzej Bialecki     <><
  ___. ___ ___ ___ _ _   __________________________________
[__ || __|__/|__||\/|  Information Retrieval, Semantic Web
___|||__||  \|  ||  |  Embedded Unix, System Integration  Contact: info at sigram dot com

To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message