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

Re: ldapmodify help?

At 08:54 AM 2001-10-11, oberwetter, josh wrote:
>I'm hoping that somebody can provide me with good advice about how to
>simplify the following process:

Don't use ldapmodify(1), use slurpd(8).  That is, treat your
Oracle DB as the master and write a program which generates
a replication log that slurpd(8) can use update slave slapd(8)s.

>The data in my directory is derived from the tables in a large Oracle
>database. Periodically a new data set is fed into Oracle, and
>changes/additions must be propagated to the directory.
>This seems fairly simple: create two different LDIFs, one that contains
>_new_ entries (so that I can use ldapmodify with the -a flag) and another
>that contains _modified_ entries (so that I can use ldapmodify with the -r
>The complication is that some of the new entries in the database may have
>been manually added to the directory between data feeds, and these additions
>woudl cause "ldapmodify -a" to throw the old "ldap_add: Already exists"
>error. Similarly, some of the new database entries have _not_ been added to
>the directory, so "ldapmodify -r" would throw a "ldap_modify: No such
>object" error.
>Does anybody know an easy way to combine modifies and adds and still use the
>alternative input format? I can write something that distinguishes betwee
>changetypes, but I'm short on time now and would prefer a quick fix.
>Will using the -c flag (continuous operation mode) cause a data integrity
>problem? Once I tried running ldapadd with one duplicate record. As expected
>an "Already exists" error was thrown and ignored, but after this I couldn't
>index the directory properly -- some attributetypes didn't index at all.