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

Re: Slapd Error Handling Problem

At 01:32 PM 9/27/99 +0200, Lars Uffmann wrote:
>After a month in production, we now had the first real problem
>with a corrupted attribute index file. Openldap version is 1.2.6 running
>on linux 2.2.10, using BerkleyDB 2:
>conn=157146 op=1 SRCH base="DC=MAIN-RHEINER,DC=DE" scope=2
>ldbm_db_errcall(): ==> /usr/local/slapd-web1/data/back-ldbm/mail.dbb:
>page 1078871380 doesn't exist, create flag not set
>conn=157146 op=1 RESULT err=0 tag=101 nentries=0
>conn=157146 op=-1 fd=21 closed errno=0

Was the index built (using ldbmindex or ldif2ldbm) after adding
"index mail ..." to your configuration file?

>What I am really concerned about is not the ldbm error, since
>we are able to rebuild the database using ldif backups.
>Slapd sends an LDAP_SUCCESS instead of eg LDAP_OPERATIONAL_ERROR back
>to the client.

That's a bug.  Submit an ITS (for this issue specifically).

>As a result of this, all email for the above user bounced
>with a 'User unknown` error. This behaviour is simply not acceptable
>since it directly affects quality of service for our users.
>Is this problem known or fixed in 1.2.7 ? I will soon upgrade to 1.2.7
>since it fixes many other open issues. Meanwhile I will have a look
>at slap/search.c and slapd/back-ldbm/search.c.

Other users are reporting this problem when using OpenLDAP 1.2.7
and BDB 2.7.7.  I have yet to determine the cause.  Feel free
to submit an ITS for this issue (page 1078871380 doesn't exist,
create flag not set) as well.