[Pkg-openldap-devel] Bug#673038: Bug#673038: slapd: slapcat output truncated every now and then

[Axel Beckert]

Hi Steve,

Steve Langasek wrote:
> > According to the slapcat man page it should be "always safe to run
> > slapcat with the slapd-bdb(5) ... backends" even if slapd runs. We do
> > use a BDB backend.
> 
> Note that the HDB backend is the one recommended upstream and the Debian
> default.

Well, yeah, that system has been dist-upgraded from at least Etch.
IIRC it started at some time when BDB was still the default.

I wrote that -- according to our backups -- this happened already with
Lenny's slapd. But with Lenny it seemed to have happened less often
(which is why we noticed it only recently).

OTOH we have far more objects in the LDAP for a while now, so it's not
clear if that increase was caused by the dist-upgrade or by going from
approximately 5000 objects to approximately 10000 objects.

> Is this issue reproducible with HDB, or only with BDB?

We will check that in a VM.

> Can you reproduce this issue with the slapd 2.4.28 package in wheezy?

We will check that, too.

> Are there any noteworthy commonalities in the entries that are being
> dropped?

It looks as if always all entries created after some time are dropped,
i.e. more recently created user were affected more often than old
accounts.

> Or perhaps commonalities in the last entries that are being
> printed?

We noticed no commonalities besides the fact that recently created
users are affected more often (the slapcat output seems to be order by
creation time). The place where it stops to output stuff seems to be
random.

> (Assumes that there is some consistent ordering to the output,
> which of course may not be the case)

>From what I have in my mind it looked quite consistently ordered.

> Would you be able to provide a minimal test configuration (/etc/ldap/slapd.d
> and LDIF of a directory, plus BDB DB_CONFIG file) that could be used to
> reproduce this?

Unfortunately not (yet) because so far the only place where we
encountered this were our live user database. But we will try to
generate some similar artifical data and check if we can reproduce
this issue with that, too.

		Regards, Axel
-- 
 ,''`.  |  Axel Beckert <abe at debian.org>, http://people.debian.org/~abe/
: :' :  |  Debian Developer, ftp.ch.debian.org Admin
`. `'   |  1024D: F067 EA27 26B9 C3FC 1486  202E C09E 1D89 9593 0EDE
  `-    |  4096R: 2517 B724 C5F6 CA99 5329  6E61 2FF9 CD59 6126 16B5