LDAP allows you to have a master+slave server setup, with automatic failover in case of problems or just to spread the load. This has obvious advantages, but it can hide problems with the master server.
Tip of the Trade: Finding and fixing master server issues isn’t always easy, especially when the slave servers seem to be working fine. With careful searching, ldapsearch handily resolves issues.
If your master server is down but the slaves are OK, ldapsearch and other lookups will work fine. But if you try to modify the database (e.g., with ldapadd), you’ll get this kind of error message:
ldap_add: Referral (10) referrals: ldaps://masterldap.example.com/uid=test,ou=People, dc=example,dc=com
Slave servers can’t modify the database, so it tries to redirect to the master, and fails. If you now use ldapsearch to look for your new entry, you will not find it.
You can confirm the diagnosis by trying: ldapsearch -H ldaps://masterldap.example.com to force a bind to the master server.
The next stage is to find out what’s causing the problem. Set the loglevel in /etc/ldap/slapd.conf to 1, then restart slapd and check the logs.
Two common problems:
- An old slapd process that hasn’t been killed off properly. Check with ps and kill it if necessary.
- An alockproblem, which looks like this in the log:
slapd: bdb_db_open: alock package is unstable slapd: backend_startup_one: bi_db_open failed! (-1)
To resolve, remove the /var/lib/ldap/alock file (or check your slapd.conf for your local data directory). Run db_recover to fix the database, or just restart slapd and it should recover itself. You may need to give it a little time or restart it again.
Remember to return the loglevel to normal afterward, or the server will be very slow. Also, consider using monitoring software to catch these silent problems!