C. Common errors encountered when using OpenLDAP Software

The following sections attempt to summarize the most common causes of LDAP errors when using OpenLDAP

C.1. Common causes of LDAP errors

C.1.1. ldap_*: Can't contact LDAP server

The Can't contact LDAP server error is usually returned when the LDAP server cannot be contacted. This may occur for many reasons:

the LDAP server is not running; this can be checked by running, for example,

      telnet <host> <port>

replacing <host> and <port> with the hostname and the port the server is supposed to listen on.

the client has not been instructed to contact a running server; with OpenLDAP command-line tools this is accomplished by providing the -H switch, whose argument is a valid LDAP url corresponding to the interface the server is supposed to be listening on.

C.1.2. ldap_*: No such object

The no such object error is generally returned when the target DN of the operation cannot be located. This section details reasons common to all operations. You should also look for answers specific to the operation (as indicated in the error message).

The most common reason for this error is non-existence of the named object. First, check for typos.

Also note that, by default, a new directory server holds no objects (except for a few system entries). So, if you are setting up a new directory server and get this message, it may simply be that you have yet to add the object you are trying to locate.

The error commonly occurs because a DN was not specified and a default was not properly configured.

If you have a suffix specified in slapd.conf eg.

      suffix "dc=example,dc=com"

You should use

      ldapsearch -b 'dc=example,dc=com' '(cn=jane*)'

to tell it where to start the search.

The -b should be specified for all LDAP commands unless you have an ldap.conf(5) default configured.

See ldapsearch(1), ldapmodify(1)

Also, slapadd(8) and its ancillary programs are very strict about the syntax of the LDIF file.

Some liberties in the LDIF file may result in an apparently successful creation of the database, but accessing some parts of it may be difficult.

One known common error in database creation is putting a blank line before the first entry in the LDIF file. There must be no leading blank lines in the LDIF file.

It is generally recommended that ldapadd(1) be used instead of slapadd(8) when adding new entries your directory. slapadd(8) should be used to bulk load entries known to be valid.

Another cause of this message is a referral ({SECT:Constructing a Distributed Directory Service}}) entry to an unpopulated directory.

Either remove the referral, or add a single record with the referral base DN to the empty directory.

This error may also occur when slapd is unable to access the contents of its database because of file permission problems. For instance, on a Red Hat Linux system, slapd runs as user 'ldap'. When slapadd is run as root to create a database from scratch, the contents of /var/lib/ldap are created with user and group root and with permission 600, making the contents inaccessible to the slapd server.

C.1.3. ldap_*: Can't chase referral

This is caused by the line

      referral        ldap://root.openldap.org

In slapd.conf, it was provided as an example for how to use referrals in the original file. However if your machine is not permanently connected to the Internet, it will fail to find the server, and hence produce an error message.

To resolve, just place a # in front of line and restart slapd or point it to an available ldap server.

See also: ldapadd(1), ldapmodify(1) and slapd.conf(5)

C.1.4. ldap_*: server is unwilling to perform

slapd will return an unwilling to perform error if the backend holding the target entry does not support the given operation.

The password backend is only willing to perform searches. It will return an unwilling to perform error for all other operations.

C.1.5. ldap_*: Insufficient access

This error occurs when server denies the operation due to insufficient access. This is usually caused by binding to a DN with insufficient privileges (or binding anonymously) to perform the operation.

You can bind as the rootdn/rootpw specified in slapd.conf(5) to gain full access. Otherwise, you must bind to an entry which has been granted the appropriate rights through access controls.

C.1.6. ldap_*: Invalid DN syntax

The target (or other) DN of the operation is invalid. This implies that either the string representation of the DN is not in the required form, one of the types in the attribute value assertions is not defined, or one of the values in the attribute value assertions does not conform to the appropriate syntax.

C.1.7. ldap_*: Referral hop limit exceeded

This error generally occurs when the client chases a referral which refers itself back to a server it already contacted. The server responds as it did before and the client loops. This loop is detected when the hop limit is exceeded.

This is most often caused through misconfiguration of the server's default referral. The default referral should not be itself:

That is, on ldap://myldap/ the default referral should not be ldap://myldap/ (or any hostname/ip which is equivalent to myldap).

C.1.8. ldap_*: operations error

In some versions of slapd(8), operationsError was returned instead of other.

C.1.9. ldap_*: other error

The other result code indicates an internal error has occurred. While the additional information provided with the result code might provide some hint as to the problem, often one will need to consult the server's log files.

C.1.10. ldap_add/modify: Invalid syntax

This error is reported when a value of an attribute does not conform to syntax restrictions. Additional information is commonly provided stating which value of which attribute was found to be invalid. Double check this value and other values (the server will only report the first error it finds).

Common causes include:

extraneous whitespace (especially trailing whitespace)
improperly encoded characters (LDAPv3 uses UTF-8 encoded Unicode)
empty values (few syntaxes allow empty values)

For certain syntax, like OBJECT IDENTIFIER (OID), this error can indicate that the OID descriptor (a "short name") provided is unrecognized. For instance, this error is returned if the objectClass value provided is unrecognized.

C.1.11. ldap_add/modify: Object class violation

This error is returned with the entry to be added or the entry as modified violates the object class schema rules. Normally additional information is returned the error detailing the violation. Some of these are detailed below.

Violations related to the entry's attributes:

      Attribute not allowed

A provided attribute is not allowed by the entry's object class(es).

      Missing required attribute

An attribute required by the entry's object class(es) was not provided.

Violations related to the entry's class(es):

      Entry has no objectClass attribute

The entry did not state which object classes it belonged to.

      Unrecognized objectClass

One (or more) of the listed objectClass values is not recognized.

      No structural object class provided

None of the listed objectClass values is structural.

      Invalid structural object class chain

Two or more structural objectClass values are not in same structural object class chain.

      Structural object class modification

Modify operation attempts to change the structural class of the entry.

      Instantiation of abstract objectClass.

An abstract class is not subordinate to any listed structural or auxiliary class.

      Invalid structural object class

C.1.12. ldap_add: No such object

The "ldap_add: No such object" error is commonly returned if parent of the entry being added does not exist. Add the parent entry first...

For example, if you are adding "cn=bob,dc=domain,dc=com" and you get:

      ldap_add: No such object

The entry "dc=domain,dc=com" likely doesn't exist. You can use ldapsearch to see if does exist:

      ldapsearch -b 'dc=domain,dc=com' -s base '(objectclass=*)'

If it doesn't, add it. See A Quick-Start Guide for assistance.

Note: if the entry being added is the same as database suffix, it's parent isn't required. i.e.: if your suffix is "dc=domain,dc=com", "dc=com" doesn't need to exist to add "dc=domain,dc=com".

This error will also occur if you try to add any entry that the server is not configured to hold.

For example, if your database suffix is "dc=domain,dc=com" and you attempt to add "dc=domain2,dc=com", "dc=com", "dc=domain,dc=org", "o=domain,c=us", or an other DN in the "dc=domain,dc=com" subtree, the server will return a "No such object" (or referral) error.

slapd(8) will generally return "no global superior knowledge" as additional information indicating its return noSuchObject instead of a referral as the server is not configured with knowledge of a global superior server.

C.1.13. ldap add: invalid structural object class chain

This particular error refers to the rule about STRUCTURAL objectclasses, which states that an object is of one STRUCTURAL class, the structural class of the object. The object is said to belong to this class, zero or more auxiliaries classes, and their super classes.

While all of these classes are commonly listed in the objectClass attribute of the entry, one of these classes is the structural object class of the entry. Thus, it is OK for an objectClass attribute to contain inetOrgPerson, organizationalPerson, and person because they inherit one from another to form a single super class chain. That is, inetOrgPerson SUPs organizationPerson SUPs person. On the other hand, it is invalid for both inetOrgPerson and account to be listed in objectClass as inetOrgPerson and account are not part of the same super class chain (unless some other class is also listed with is a subclass of both).

To resolve this problem, one must determine which class will better serve structural object class for the entry, adding this class to the objectClass attribute (if not already present), and remove any other structural class from the entry's objectClass attribute which is not a super class of the structural object class.

Which object class is better depends on the particulars of the situation. One generally should consult the documentation for the applications one is using for help in making the determination.

C.1.14. ldap_add: no structuralObjectClass operational attribute

ldapadd(1) may error:

      adding new entry "uid=XXX,ou=People,o=campus,c=ru"
        ldap_add: Internal (implementation specific) error (80)
           additional info: no structuralObjectClass operational attribute

when slapd(8) cannot determine, based upon the contents of the objectClass attribute, what the structural class of the object should be.

C.1.15. ldap_add/modify/rename: Naming violation

OpenLDAP's slapd checks for naming attributes and distinguished values consistency, according to RFC 4512.

Naming attributes are those attributeTypes that appear in an entry's RDN; distinguished values are the values of the naming attributes that appear in an entry's RDN, e.g, in

      cn=Someone+mail=someone@example.com,dc=example,dc=com

the naming attributes are cn and mail, and the distinguished values are Someone and someone@example.com.

OpenLDAP's slapd checks for consistency when:

adding an entry
modifying an entry, if the values of the naming attributes are changed
renaming an entry, if the RDN of the entry changes

Possible causes of error are:

the naming attributes are not present in the entry; for example:

                dn: dc=example,dc=com
                objectClass: organization
                o: Example
                # note: "dc: example" is missing

the naming attributes are present in the entry, but in the attributeType definition they are marked as:
- collective
- operational
- obsolete
the naming attributes are present in the entry, but the distinguished values are not; for example:

                dn: dc=example,dc=com
                objectClass: domain
                dc: foobar
                # note: "dc" is present, but the value is not "example"

the naming attributes are present in the entry, with the distinguished values, but the naming attributes:
- do not have an equality field, so equality cannot be asserted
- the matching rule is not supported (yet)
- the matching rule is not appropriate
the given distinguished values do not comply with their syntax
other errors occurred during the validation/normalization/match process; this is a catchall: look at previous logs for details in case none of the above apply to your case.

In any case, make sure that the attributeType definition for the naming attributes contains an appropriate EQUALITY field; or that of the superior, if they are defined based on a superior attributeType (look at the SUP field). See RFC 4512 for details.

C.1.16. ldap_add/delete/modify/rename: no global superior knowledge

If the target entry name places is not within any of the databases the server is configured to hold and the server has no knowledge of a global superior, the server will indicate it is unwilling to perform the operation and provide the text "no global superior knowledge" as additional text.

Likely the entry name is incorrect, or the server is not properly configured to hold the named entry, or, in distributed directory environments, a default referral was not configured.

C.1.17. ldap_bind: Insufficient access

Current versions of slapd(8) requires that clients have authentication permission to attribute types used for authentication purposes before accessing them to perform the bind operation. As all bind operations are done anonymously (regardless of previous bind success), the auth access must be granted to anonymous.

In the example ACL below grants the following access:

to anonymous users:
- permission to authenticate using values of userPassword
to authenticated users:
- permission to update (but not read) their userPassword
- permission to read any object excepting values of userPassword

All other access is denied.

        access to attr=userPassword
          by self =w
          by anonymous auth
        access *
          by self write
          by users read

C.1.18. ldap_bind: Invalid credentials

The error usually occurs when the credentials (password) provided does not match the userPassword held in entry you are binding to.

The error can also occur when the bind DN specified is not known to the server.

Check both! In addition to the cases mentioned above you should check if the server denied access to userPassword on selected parts of the directory. In fact, slapd always returns "Invalid credentials" in case of failed bind, regardless of the failure reason, since other return codes could reveal the validity of the user's name.

To debug access rules defined in slapd.conf, add "ACL" to log level.

C.1.19. ldap_bind: Protocol error

There error is generally occurs when the LDAP version requested by the client is not supported by the server.

The OpenLDAP Software 2.x server, by default, only accepts version 3 LDAP Bind requests but can be configured to accept a version 2 LDAP Bind request.

Note: The 2.x server expects LDAPv3 [RFC4510] to be used when the client requests version 3 and expects a limited LDAPv3 variant (basically, LDAPv3 syntax and semantics in an LDAPv2 PDUs) to be used when version 2 is expected.

This variant is also sometimes referred to as LDAPv2+, but differs from the U-Mich LDAP variant in a number of ways.

C.1.20. ldap_modify: cannot modify object class

This message is commonly returned when attempting to modify the objectClass attribute in a manner inconsistent with the LDAP/X.500 information model. In particular, it commonly occurs when one tries to change the structure of the object from one class to another, for instance, trying to change an 'apple' into a 'pear' or a 'fruit' into a 'pear'.

Such changes are disallowed by the slapd(8) in accordance with LDAP and X.500 restrictions.

C.1.21. ldap_sasl_interactive_bind_s: ...

If you intended to bind using a DN and password and get an error from ldap_sasl_interactive_bind_s, you likely forgot to provide a '-x' option to the command. By default, SASL authentication is used. '-x' is necessary to select "simple" authentication.

C.1.22. ldap_sasl_interactive_bind_s: No such Object

This indicates that LDAP SASL authentication function could not read the Root DSE. The error will occur when the server doesn't provide a root DSE. This may be due to access controls.

C.1.23. ldap_sasl_interactive_bind_s: No such attribute

This indicates that LDAP SASL authentication function could read the Root DSE but it contained no supportedSASLMechanism attribute.

The supportedSASLmechanism attribute lists mechanisms currently available. The list may be empty because none of the supported mechanisms are currently available. For example, EXTERNAL is listed only if the client has established its identity by authenticating at a lower level (e.g. TLS).

Note: the attribute may not be visible due to access controls

Note: SASL bind is the default for all OpenLDAP tools, e.g. ldapsearch(1), ldapmodify(1). To force use of "simple" bind, use the "-x" option. Use of "simple" bind is not recommended unless one has adequate confidentiality protection in place (e.g. TLS/SSL, IPSEC).

C.1.24. ldap_sasl_interactive_bind_s: Unknown authentication method

This indicates that none of the SASL authentication supported by the server are supported by the client, or that they are too weak or otherwise inappropriate for use by the client. Note that the default security options disallows the use of certain mechanisms such as ANONYMOUS and PLAIN (without TLS).

Note: SASL bind is the default for all OpenLDAP tools. To force use of "simple" bind, use the "-x" option. Use of "simple" bind is not recommended unless one has adequate confidentiality protection in place (e.g. TLS/SSL, IPSEC).

C.1.25. ldap_sasl_interactive_bind_s: Local error (82)

Apparently not having forward and reverse DNS entries for the LDAP server can result in this error.

C.1.26. ldap_search: Partial results and referral received

This error is returned with the server responses to an LDAPv2 search query with both results (zero or more matched entries) and references (referrals to other servers). See also: ldapsearch(1).

If the updatedn on the replica does not exist, a referral will be returned. It may do this as well if the ACL needs tweaking.

C.1.27. ldap_start_tls: Operations error

ldapsearch(1) and other tools will return

        ldap_start_tls: Operations error (1)
              additional info: TLS already started

When the user (though command line options and/or ldap.conf(5)) has requested TLS (SSL) be started twice. For instance, when specifying both "-H ldaps://server.do.main" and "-ZZ".

C.2. Other Errors

C.2.1. ber_get_next on fd X failed errno=34 (Numerical result out of range)

This slapd error generally indicates that the client sent a message that exceeded an administrative limit. See sockbuf_max_incoming and sockbuf_max_incoming_auth configuration directives in slapd.conf(5).

C.2.2. ber_get_next on fd X failed errno=11 (Resource temporarily unavailable)

This message is not indicative of abnormal behavior or error. It simply means that expected data is not yet available from the resource, in this context, a network socket. slapd(8) will process the data once it does becomes available.

C.2.3. daemon: socket() failed errno=97 (Address family not supported)

This message indicates that the operating system does not support one of the (protocol) address families which slapd(8) was configured to support. Most commonly, this occurs when slapd(8) was configured to support IPv6 yet the operating system kernel wasn't. In such cases, the message can be ignored.

C.2.4. GSSAPI: gss_acquire_cred: Miscellaneous failure; Permission denied;

This message means that slapd is not running as root and, thus, it cannot get its Kerberos 5 key from the keytab, usually file /etc/krb5.keytab.

A keytab file is used to store keys that are to be used by services or daemons that are started at boot time. It is very important that these secrets are kept beyond reach of intruders.

That's why the default keytab file is owned by root and protected from being read by others. Do not mess with these permissions, build a different keytab file for slapd instead, and make sure it is owned by the user that slapd runs as.

To do this, start kadmin, and enter the following commands:

     addprinc -randkey ldap/ldap.example.com@EXAMPLE.COM
     ktadd -k /etc/openldap/ldap.keytab ldap/ldap.example.com@EXAMPLE.COM

Then, on the shell, do:

     chown ldap:ldap /etc/openldap/ldap.keytab
     chmod 600 /etc/openldap/ldap.keytab

Now you have to tell slapd (well, actually tell the gssapi library in Kerberos 5 that is invoked by Cyrus SASL) where to find the new keytab. You do this by setting the environment variable KRB5_KTNAME like this:

     export KRB5_KTNAME="FILE:/etc/openldap/ldap.keytab"

Set that environment variable on the slapd start script (Red Hat users might find /etc/sysconfig/ldap a perfect place).

This only works if you are using MIT kerberos. It doesn't work with Heimdal, for instance.

In Heimdal there is a function gsskrb5_register_acceptor_identity() that sets the path of the keytab file you want to use. In Cyrus SASL 2 you can add

    keytab: /path/to/file

to your application's SASL config file to use this feature. This only works with Heimdal.

C.2.5. access from unknown denied

This related to TCP wrappers. See hosts_access(5) for more information. in the log file: "access from unknown denied" This related to TCP wrappers. See hosts_access(5) for more information. for example: add the line "slapd: .hosts.you.want.to.allow" in /etc/hosts.allow to get rid of the error.

C.2.6. ldap_read: want=# error=Resource temporarily unavailable

This message occurs normally. It means that pending data is not yet available from the resource, a network socket. slapd(8) will process the data once it becomes available.

C.2.7. `make test' fails

Some times, `make test' fails at the very first test with an obscure message like

    make test
    make[1]: Entering directory `/ldap_files/openldap-2.5.0/tests'
    make[2]: Entering directory `/ldap_files/openldap-2.5.0/tests'
    Initiating LDAP tests for MDB...
    Cleaning up test run directory leftover from previous run.
     Running ./scripts/all...
    >>>>> Executing all LDAP tests for mdb
    >>>>> Starting test000-rootdse ...
    running defines.sh
    Starting slapd on TCP/IP port 9011...
    Using ldapsearch to retrieve the root DSE...
    Waiting 5 seconds for slapd to start...
    ./scripts/test000-rootdse: line 40: 10607 Segmentation fault $SLAPD -f $CONF1 -h $URI1 -d $LVL $TIMING >$LOG1 2>&1
    Waiting 5 seconds for slapd to start...
    Waiting 5 seconds for slapd to start...
    Waiting 5 seconds for slapd to start...
    Waiting 5 seconds for slapd to start...
    Waiting 5 seconds for slapd to start...
    ./scripts/test000-rootdse: kill: (10607) - No such pid
    ldap_sasl_bind_s: Can't contact LDAP server (-1)
    >>>>> Test failed
    >>>>> ./scripts/test000-rootdse failed (exit 1)
    make[2]: *** [mdb-yes] Error 1
    make[2]: Leaving directory `/ldap_files/openldap-2.5.0/tests'
    make[1]: *** [test] Error 2
    make[1]: Leaving directory `/ldap_files/openldap-2.5.0/tests'
    make: *** [test] Error 2

or so. Usually, the five lines

Waiting 5 seconds for slapd to start...

indicate that slapd didn't start at all.

In tests/testrun/slapd.1.log there is a full log of what slapd wrote while trying to start. The log level can be increased by setting the environment variable SLAPD_DEBUG to the corresponding value; see loglevel in slapd.conf(5) for the meaning of log levels.

A typical reason for this behavior is a runtime link problem, i.e. slapd cannot find some dynamic libraries it was linked against. Try running ldd(1) on slapd (for those architectures that support runtime linking).

There might well be other reasons; the contents of the log file should help clarifying them.

Tests that fire up multiple instances of slapd typically log to tests/testrun/slapd.<n>.log, with a distinct <n> for each instance of slapd; list tests/testrun/ for possible values of <n>.

C.2.8. ldap_*: Internal (implementation specific) error (80) - additional info: entry index delete failed

This seems to be related with wrong ownership of the MDB's dir (/var/lib/ldap) and files. The files must be owned by the user that slapd runs as.

    chown -R ldap:ldap /var/lib/ldap

fixes it in Debian

C.2.9. ldap_sasl_interactive_bind_s: Can't contact LDAP server (-1)

Using SASL, when a client contacts LDAP server, the slapd service dies immediately and client gets an error :

     SASL/GSSAPI authentication started ldap_sasl_interactive_bind_s: Can't contact LDAP server (-1)

Then check the slapd service, it stopped.