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

Re: alock + ldbm in 2.3.21

On Wed, 2006-04-12 at 16:18 -0400, matthew sporleder wrote:

> I wasn't aware that LDBM was susceptible to the same inconsistencies
> as BDB.  I thought the lack of transactions/checkpointing prevented
> this.  And if this situation did occur, I thought 2.3 was smart enough
> to try fixing itself.

Sounds like you're reversing things: ldbm is subject to the same risk of
inconsistencies of any other software component that relies on
BerkeleyDB (no software is perfect).  In fact, they use the same storage
system but, since back-bdb (and back-hdb) exploit many features of
BerkeleyDB that back-ldbm doesn't, they offer bth better performances
and more reliability/recoverability.  The fact that back-ldbm happily
gets corrupted without letting you know seems to be a drawback rather
than a feature.

> I did see some evidence of corruption, but I thought it was due to a
> problem with db_upgrade, or a misconfiguration.  Nevertheless, I will
> try a db_recover.
> (dn2entry_r: no entry for valid id (5) dn...)

I don't remember the details, but I'm pretty sure there's no binary
compatibility between database files created by OpenLDAP 2.1 and
OpenLDAP 2.3; in general, you shouldn't expect anything like that when
moving across minor versions.  As Quanah already pointed out, migration
from minor to minor should be done via (old)slapcat/(new)slapadd, which,
when using OpenLDAP 2.3, exploits the very good performances of slapadd.


Ing. Pierangelo Masarati
Responsabile Open Solution
OpenLDAP Core Team

SysNet s.n.c.
Via Dossi, 8 - 27100 Pavia - ITALIA
Office:   +39.02.23998309          
Mobile:   +39.333.4963172
Email:    pierangelo.masarati@sys-net.it