On 27/08/2009 23:17, Serge Dubrouski wrote:
Hello -
In our infrastructure we have one master and 2 shadow OpenLDAP 2.4.13 servers with db-4.6.21 backend database, The directory contains about 110000 records. Today we had to delete about 2000 records on the master and that caused a complete refresh on shadows. I mean shadows started to delete all records in their databases. In the log files it started with appearing these messages:
Aug 27 14:59:40 berlin slapd[4138]: do_syncrep2: rid=011 LDAP_RES_INTERMEDIATE - SYNC_ID_SET Aug 27 14:59:40 berlin slapd[4138]: do_syncrep2: cookie=rid=011,csn=20090827205939.642722Z#000000#000#000000
Then it started deleting records:
Aug 27 14:59:45 berlin slapd[4138]: syncrepl_del_nonpresent: rid=011 be_delete cn=,,,,, Aug 27 14:59:46 berlin slapd[4138]: syncrepl_del_nonpresent: rid=011 be_delete cn=..... Aug 27 14:59:46 berlin slapd[4138]: syncrepl_del_nonpresent: rid=011 be_delete cn=..... Aug 27 14:59:46 berlin slapd[4138]: syncrepl_del_nonpresent: rid=011 be_delete cn=.... Aug 27 14:59:46 berlin slapd[4138]: syncrepl_del_nonpresent: rid=011 be_delete cn=.... Aug 27 14:59:46 berlin slapd[4138]: syncrepl_del_nonpresent: rid=011 be_delete cn=...
And deleted records that were deleted records for a whole directory and then started putting them back. Does anybody know if this is expected behavior?
Hi Serge,
OpenLDAP version 2.4.13 came pretty early in the 2.4 lifecycle, and as such contained bugs like the one you're referring to. It is a pretty old version, nearly a year back now!
Many bugs around replication have been corrected since then. I do recommend you move on to 2.4.17, the latest version, which will likely not have such behaviour. I use it in multi-master and master-slave configurations on large databases and it "just works" :-)
Regards, Jonathan