Hello, list.

I'm experiencing synchronization problem with my ldap provider-consumer setup.
Both nodes are running OpenLDAP 2.4.38 (using mdb backend) at x86_64
under Linux 3.7.10.

Both servers synchronize their time with one NTP server, so clocks are in sync.

I have cn=dhcp,dc=my,dc=org containter which I would like to replicate to openldap

Consumer uses old-fasioned slapd.conf file:
# [... include, acl and logging settings skipped ...]

rootdn        "cn=root,dc=my,dc=org"
rootpw        [ skipped ]

index    objectClass    eq
index    entryCSN    eq
index    entryUUID    eq
index    dlzHostname    eq
index    dlzZoneName    eq
index    dlzIPAddr    eq
index    dlzType        eq
index    dhcpHWAddress    eq
index    cn        eq,approx,sub

syncrepl    rid=1
syncrepl    rid=2
    retry="60 +"
    credentials="[ skipped ]"

Problem is that not all changes (adding new objects and changing attributes of existing
objects) are replicated from master to consumer.  I make change on master by hand, but
I do not see log entries on consumer that changed attribute was replicated.

LDAP provider setup (uses online configuartion):
dn: olcDatabase={1}mdb,cn=config
objectClass: olcDatabaseConfig
objectClass: olcMdbConfig
olcDatabase: {1}mdb
olcDbDirectory: /var/lib/ldap
olcSuffix: dc=my,dc=org
olcAddContentAcl: FALSE
olcLastMod: TRUE
olcLimits: {0}group/groupOfNames/member="cn=ldap admins,ou=groups,dc=my,dc=org" size=unlimited
olcLimits: {1}group/groupOfNames/member="cn=ldap admins,ou=groups,dc=my,dc=org" time=unlimited
olcLimits: {2}group/groupOfNames/member="cn=admins,ou=mail,ou=groups,dc=my,dc=org" size=unlimited
olcLimits: {3}group/groupOfNames/member="cn=replicators,ou=groups,dc=my,dc=org" size=unlimited time=unlimited

User uid=dhcpd,ou=services,dc=my,dc=org is member of group cn=replicators,ou=groups,dc=my,dc=org.
dn: cn=Replicators,ou=Groups,dc=my,dc=org
cn: Replicators
member: uid=dhcpd,ou=services,dc=my,dc=org

Indices on master server:
olcDbIndex: dhcpHWAddress,dhcpClassData eq

ACL for cn=dhcp,dc=my,dc=org container (I have only one ACL entry, that mentions cn=dhcp,dc=my,dc=org):
olcAccess: {15}to dn.subtree="cn=dhcp,dc=my,dc=org"
  by group/groupOfNames/member.exact="cn=dhcp readers,ou=dhcp,ou=groups,dc=my,dc=org" read
  by group/groupOfNames/member.exact="cn=dhcp writers,ou=dhcp,ou=groups,dc=my,dc=org" write

dn: cn=dhcp readers,ou=dhcp,ou=Groups,dc=my,dc=org
cn: dhcp readers
objectClass: groupOfNames
objectClass: top
member: uid=dhcpd,ou=Services,dc=my,dc=org

So for now I checked following:
1. Clocks on both server (both are in sync)
2. ACL for replication DN (replication user is able to read all needed data).
3. Size and time limits for replication DN (replication user has no limits on operations).
4. Schema (objectClasses and attrubutes) on both servers (both nodes use same schemas).
5. Disable indices on consumer server for replicated attributes from cn=dhcp,dc=my,dc=org container (not helped)

What else I can check?

BTW, I would be pleased if someone give somekind of systematic approach for troubleshooting OpenLDAP replication issues.