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

Re: Data portability between slaves

--On Friday, March 31, 2006 12:48 PM +1100 Dave Horsfall <daveh@ci.com.au> wrote:

On Thu, 30 Mar 2006, Quanah Gibson-Mount wrote:

> Given two slaves with identical releases and data, how portable are
> the BDB files and SLAPCAT files between them?  That is, does the
> entryUUID attribute (or anything else) contain host-specific
> information?

They are identical (or should be).


I've recovered a slave after hard drive failure several times simply by
stopping slapd on a different slave, and then copying the database from
that replica to the one with the new hard drive (note that you want to
run db_recover after stopping slapd, before copying over the files).

To make db_recover update the *.bdb files from the log? I know it's done on startup, but I would've thought that a sync would have been done on shutdown, too. That explains a couple of things I've seen...

To force the system to checkpoint (flush the BDB environment).

entryUUID, etc, *must* be the same on all systems for syncrepl to work

Yes, but I was concerned that each host would have its own fingerprint somewhere...

You need to grab the BDB log files, too (just to be clear).


-- Quanah Gibson-Mount Principal Software Developer ITS/Shared Application Services Stanford University GnuPG Public Key: http://www.stanford.edu/~quanah/pgp.html