[Date Prev][Date Next] [Chronological] [Thread] [Top]

RE: Index problems

Sean O'Malley wrote:
> I don't think Im the only one that gets caught offguard with 187M ldif
> file using up 4G in log files and around 500M for the data.

Yes, you aren't.

> The 500M for the data was about what I expected. The 4G in logs is
> what caught me offguard..

Here: 330M ldif, 750M data, 4G (or something like that) logs.

> I guess the only reason why I mention this is because it took 60 hours
> to get the data into the db

60 hours?
Make sure you a) turn off indexing during population and b) put the DB 
on a RAM Disk if you can.
I filled my directory in 40 minutes, indexing took another 30 minutes 
then. (4 eq indeces.)

> and im reloading it because I didn't realize the log files were that
> important if I wasn't planning on running slurp.

Well, transaction log has absolutely nothing to do with replication. 
Moreover it isn't even related to OpenLDAP, but BDB.

> Also while Im being a chronic complainer. Can someone add something to
> the documentation about the indexing features, like what they actually
> mean to the system admin trying to determine whether they are needed
> or not.

Well, there's a hint somehwere in the FAQ-O-Matic, I think, that states, 
that the type of index to be used depends on the type of searchfilter 
the software uses. Maybe someone with more knoweledge on this topic can 
comment, but basically this translates into:

index           filter
pres            "*"
eq              "foobar"
sub[str]        "*oo*a*"
  subinitial    "*bar"
... etc.

This should be sufficient for an adimistrator to configure the server. 
Though this should be mentioned in the admin guide, slapd-(bdb|ldbm) 
manpage and the FAQs.

> I think I finally figured out what I needed but I ended up
> having to look at the RFC. It would be nice if that was clearly
> explained to a dufus like me in the manual.


IE brings the web to Unix...
Isn't that like Ronald McDonald brings religion to the Pope?