Hi,

An add operation was underway on my HDB backend.  Around one second later, a large group modification ran slapd out of memory.  Slapd was then unresponsive until a watchdog killed and restarted it two minutes later.

The add operation successfully modified the database but never appeared in the audit log nor the syncrepl access log, so it was never replicated to the replicas.

Is it by design that modifications hit the database before being recorded in the audit and access logs?

This is with OpenLDAP 2.3.39, BDB 4.4.

Logs:
Sep 22 09:15:16 host slapd[27404]: conn=102591 op=5 ADD dn="uid=username,ou=people,dc=example,dc=com"
...
Sep 22 09:15:17 host slapd[27404]: conn=99395 op=131974 MOD dn="cn=biggroup,ou=groups,dc=example,dc=com"
Sep 22 09:15:17 host slapd[27404]: conn=99395 op=131974 MOD attr=member memberuid
Sep 22 09:15:17 host slapd[27404]: ch_malloc of 242464 bytes failed
Sep 22 09:15:17 host slapd[27404]: conn=102732 op=2 UNBIND
Sep 22 09:15:17 host slapd[27404]: conn=102732 fd=124 closed
Sep 22 09:17:33 host slapd: allocation failed: 0
Sep 22 09:17:33 host slapd: allocation failed: 12

Config snippet:
database hdb
suffix "dc=example,dc=com"
rootdn "cn=manager,dc=example,dc=com"
rootpw ...
directory /var/lib/ldap
cachesize 25000
idlcachesize 75000
checkpoint 512 1
lastmod on
... indexes ...
... ACLs ...

overlay auditlog
auditlog /var/lib/ldif/auditlog/example.com.ldif

overlay syncprov
syncprov-sessionlog 1000

overlay accesslog
logdb cn=accesslog
logops writes
logsuccess TRUE
logpurge 02+23:46 01+23:46

--
Thanks,
Sean Burford