Henning Brauer wrote: > * Philip Guenther <guent...@gmail.com> [2009-01-06 00:40]: >> On Mon, Jan 5, 2009 at 11:15 AM, Claudio Jeker <cje...@diehard.n-r-g.com> >> wrote: >> ... >>> Any DB that needs human help after a crash is in my opinion a bad choice. >> So that would rule out the ldbm backend, no? Last I checked the libc >> btree code, a crash while writing out a page split would corrupt the >> subtree. > > I am using openldap with ldbm backend in an not exactly small > installation for 9 or 10 years now. I have never ever experienced a > broken database. never.
I second that, 5+ years of ldbm backend usage without any problems. We've had poweroutages, disks running full, all sorts of mischief, but never a problem with corrupt ldbm databases. Ever. I *did* have thee times of huge trouble with bdb as a backend. Two times with unexpected halts of the system after which the slapd process would simply not be able to read the bdb files anymore and one time with a prickly problem where I needed to upgrade my database because bdb was updated on that system (not a major release mind you, 4.x to 4.newer_x). I know this is all not very scientific evidence regarding the stability and robustness of bdb, but I guess it is hard to forget the pain that came from using bdb in the couple of times I had to or did so unknowingly. Mind you, the last time I've used bdb is 4 years ago. Things might have changed these days. > trying bdb lead to disasters all over the place. but admittedly that > was many many many moons ago. > > openldap is still a piece of shit, but the ldbm backend is probably the > sanest one.