overview terminology

Wim Lewis wiml at omnigroup.com
Fri Jun 14 21:38:28 UTC 2002



On Friday, June 14, 2002, at 12:31  PM, Russell S. Ireland wrote:
> I am somewhat confused by some descriptions used in a few of the inn 
> man pages.
>
> "MAKEDBZ(8)
> 	Makedbz rebuilds dbz(3) database."
>
> So what is the actual datbase?  The history file?  The overview 
> directory
> structure?

The DBZ database is an index file which makes it faster to find records 
in the history file. Originally it was a hash table which allowed you to 
look up a message-ID and get the byte offset of that message's line in 
the history file without having to scan through the history file. I 
don't know if that's still how it's implemented, but AFAIK that's still 
its purpose.

> 	"WARNING: If you're trying to rebuild the overview
>        database, be sure to delete or zero out the existing
>        database before you start [....]"
>
> Again, the use of "database" seems ambiguous.  I would love to heed this
> warning, so what exactly do I delete or zero out?  Which file(s)?

This depends on what overview database method you're using --- the 
'ovmethod' entry in inn.conf. I'm using buffindexed, and so my overview 
database is stored in a few files listed in buffindexed.conf; I zero 
those out by dd'ing from /dev/zero before rebuilding my overview 
database.

So:
   - Articles are stored somewhere (in CNFS, or one-file-per-article in 
traditional methods, etc.)
   - The history file is a simple list of what articles the server knows 
it has and where they're stored.
   - The dbz database is an index for fast access to the history file.
   - The overview database (which is only needed/used if you have 
readers, not if you're purely a feed) can be stored in one of a number 
of ways and is separate from the history file.

> So if I have placed a tarball of another server's articles on my new 
> server,
> do I need to run makedbz AND makehistory to get a current history and
> overview?  Or is makehistory -O -F -f history.n sufficient?

Yes, I think you need to run makedbz as well. The example in the 
makehistory man page fits your situation pretty well --- INN's various 
indexes are out of date since you've modified the spool by hand, so you 
want it to scan the spool and rebuild the indexes from the actual 
article data.

Wim.



More information about the inn-workers mailing list