OpenLDAP Faq-O-Matic : Trash : Why is ldif2ldbm (database creation) so slow? | |
This answer is depreciated.
ldif2ldbm
(versions 1.1.2 and later) automaticatically disable per
operation cache-to-disk writing synchronization.
| |
Normally slapd's backend database
cache is sync'ed to disk on each LDAP write operations. Since creating a
database is pure writing,
ldif2ldbm
(versions 1.1.1 and prior)
has to sync every entry individually. To keep this from happening, add the following (on a line of its own) to your slapd.conf. After you have created the database you should comment out this line by inserting a # in front of the option.dbCacheNoWSync
This can make an incredible speed difference
(eg. 2000 entries can take about 3 hours with sync on
or 3 minutes with sync off)!
| |
Use dbcacheNoWsync with care. Disabling cache-to-disk
write synchronization may result is lost of data or database corruption
when slapd is abnormally
terminated.
| |
Another trick is to create your database on a virtual memory
filesystem (MFS, tmpfs, or whatever) and then move the database
to a disk filesystem (and changing slapd.conf appropriately) before
cranking up slapd. This will significantly decrease the time needed to create the database dramatically. | |
[Append to This Answer] |
|