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

Re: (ITS#4264) test033-glue-syncrepl with hdb: comparison failed - database was not created correctly



On Fri, 2005-12-16 at 10:16 +0000, michael@stroeder.com wrote:
> Pierangelo Masarati wrote:
> > 
> > In the meanwhile, I managed to reproduce it (randomly, and very seldom).
> > [..]
> > SLAPD_DEBUG=stats,stats2,trace,args,sync ./run -b hdb test033
> 
> Hmm, it often runs successfully when using this debug level. It always
> fails on my system when just invoking ./run -b hdb test033.
> 
> Maybe something related to timing?

I think I got somewhere.  It shouldn't be a bug per se, it rather has to
do with the way the databases are generated.  The setup is made of:

- a common root
  - a branch that's provider in 1 and consumer in 2
  - a branch that's consumer in 1 and provider in 2

to generate the database, I recycled the slapd.conf of test011 &
test012, added the entire database with slapadd, then I copied the
common database directory, assigned the first branch to 1 and the second
to 2 and relied on refresh to load the missing parts.  This might have
created some issue with generated CSNs, because it happens, every now
and then, that entries available in the producer are skipped by the
initial syncprov search because their CSN is greater than that of the
snapshot.  This occasional flawed behavior wasn't present from the
beginning, I think it started when snapshotting was introduced in
syncprov.  Now I'm entirely reworking the database setup, and in fact I
couldn't see any more failures (but I better keep testing before I claim
this).

p.




Ing. Pierangelo Masarati
Responsabile Open Solution

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