OID 11g Multimaster Replication Intermittently Fails to Update Replicas (Doc ID 1612220.1)

Last updated on APRIL 12, 2021

Applies to:

Symptoms

Oracle Internet Directory (OID) 11g.

One MDS site (e.g., <SUPPLIER_HOSTNAME>) and two RMS sites (e.g., <CONSUMER1_HOSTNAME> and <CONSUMER2_HOSTNAME>).

Replication was working in 10g 10.1.4.3, before the upgrade to 11g 11.1.1.7.

Updates are not being applied to some of the replicas intermittently.

For example:

1. Modified userPassword attribute for a user on server <CONSUMER1_HOSTNAME>
2. This change was successfully replicated to server <SUPPLIER_HOSTNAME>
3. This change was not replicated to server <SUPPLIER_HOSTNAME>

4. Modified userPassword attribute for a user on server <SUPPLIER_HOSTNAME>
5. This change was not replicated to server <CONSUMER1_HOSTNAME>
6. This change was replicated to <SUPPLIER_HOSTNAME>

Or:

1. Modified userPassword attribute for a user on server <SUPPLIER_HOSTNAME>
2. This change was successfully replicated to server <CONSUMER1_HOSTNAME>
3. This change was successfully replicated to server <SUPPLIER_HOSTNAME>

4. Modified userPassword attribute for a user on server <CONSUMER1_HOSTNAME>
5. This change was successfully replicated to server <SUPPLIER_HOSTNAME>
6. This change was not replicated to <SUPPLIER_HOSTNAME>

No errors or failures on debugged Replication or OID ldap server logs.  No occurrences of the replicated entries in the debugged OID server logs either.

A row is inserted into the ASR_CHG_LOG table on the replica that did not get updated, but the RETRY_CNT column gets updated to -2 as if completed successfully.

(Reference:
How ASR Replication Works <Document 727937.1>
How To Manage OID Replication Conflicts in the HIQ (Human Intervention Queue) Using HIQPURGE and HIQRETRY <Document 171693.1> )

If manually changing the RETRY_CNT column back to zero on the ASR_CHG_LOG row, the replica is then updated.

Tried separating OID ldap processes for Replication as per unrelated <Document 1375895.1> Solution, but the problem continues.

Testing ASR replication at the DB level, i.e., by inserting into the asr_chg_log table from all 3 servers, always works consistently.

No errors or failure from remtool checks, only from remtool -dispasrerr -v, for example:

At startup time, the OID logs show the following error:

But the replication configuration entries look correct and the same on all nodes and have not been changed since 10g (so it is not related to <Bug 5234318>).

Changes

Cause

My Oracle Support provides customers with access to over a million knowledge articles and a vibrant support community of peers and Oracle experts.