> I was having problem with syncrepl ( 2.3 and bdb 4.3 ). some of the > entries didn't go to the slave and sometimes, old entries were gone too.. > > do you mind share your configuration ( both master and slave )? >
Sure thing. I'll be doing some heavier testing in a week, so I'll let everyone know if I see any similar results. I'm also going to try larger datasets to see if problems start showing up in the 2 million entries range as suggested by Quanah and start doing modifies as well. Although my dataset personally will probably never get that high. Both the master and the two slaves are running on Dell PowerEdge 1850's. They have 2 2.8G Xeon CPUs and 2G RAM. They have two scsi disks running raid1. They are all running FreeBSD 5.4, with SMP enabled in the kernel. Both BerkeleyDB and Openldap are built from source. Openldap is RELENG_2_3 from CVS on 8/11. The RADIUS schema is taken from the FreeRadius project and slightly modified, which explains some of the attributes I'm indexing. The .lab TLD is just one I setup in our lab nameservers. slapd.conf (master) include /usr/local/openldap23/etc/openldap/schema/core.schema include /usr/local/openldap23/etc/openldap/schema/RADIUS-LDAPv3.schema #referral ldap://root.openldap.org pidfile /usr/local/openldap23/var/run/slapd.pid argsfile /usr/local/openldap23/var/run/slapd.args loglevel 256 database hdb suffix "dc=mydomain,dc=net" rootdn "cn=root,dc=mydomain,dc=net" rootpw secret directory /usr/local/openldap23/var/openldap-data # Indices to maintain index objectClass eq index uid eq index radiusGroupName eq index accountNumber eq index entryUUID eq index entryCSN eq limits dn.exact="cn=slave,ou=admins,dc=mydomain,dc=net" size=unlimited time=unlimited cachesize 100000 idlcachesize 300000 checkpoint 1024 5 overlay syncprov syncprov-checkpoint 100 10 syncprov-sessionlog 1000 access to dn.sub="dc=mydomain,dc=net" by dn.exact="cn=billing,ou=admins,dc=mydomain,dc=net" write by dn.exact="cn=slave,ou=admins,dc=mydomain,dc=net" read by self read by anonymous auth by * none database monitor access to dn.sub="cn=monitor" by dn.exact="cn=monitor,ou=admins,dc=mydomain,dc=net" read by * none slapd.conf (slave) include /usr/local/openldap23/etc/openldap/schema/core.schema include /usr/local/openldap23/etc/openldap/schema/RADIUS-LDAPv3.schema #referral ldap://root.openldap.org pidfile /usr/local/openldap23/var/run/slapd.pid argsfile /usr/local/openldap23/var/run/slapd.args loglevel 256 database hdb suffix "dc=mydomain,dc=net" rootdn "cn=root,dc=mydomain,dc=net" rootpw secret directory /usr/local/openldap23/var/openldap-data # Indices to maintain index objectClass eq index uid eq index radiusGroupName eq index accountNumber eq index entryUUID eq index entryCSN eq cachesize 100000 idlcachesize 300000 checkpoint 1024 5 syncrepl rid=1 provider=ldap://ldapm.lab:389 type=refreshAndPersist retry="60 30 300 +" searchbase="dc=mydomain,dc=net" filter="(objectclass=*)" scope=sub attrs=* sizelimit=none timelimit=none schemachecking=on bindmethod=simple binddn="cn=slave,ou=admins,dc=mydomain,dc=net" credentials="slave" updateref ldap://ldapm.lab access to dn.sub="dc=mydomain,dc=net" by dn.exact="cn=slave,ou=admins,dc=mydomain,dc=net" write by self read by anonymous auth by * none database monitor access to dn.sub="cn=monitor" by dn.exact="cn=monitor,ou=admins,dc=mydomain,dc=net" read by * none -------- I uncomment DB_TXT_NOSYNC when doing the original slapadd - then comment it back when I start for the first time. This causes slapd to run db_recover automatically to update the BDB environment and re-enable logging. DB_CONFIG (all servers) set_cachesize 0 536870912 1 set_lg_regionmax 1048576 set_lg_bsize 2097152 set_lg_max 10485760 set_flags DB_LOG_AUTOREMOVE #set_flags DB_TXN_NOSYNC
