Australian cloud for business software — built and run in Sydney since 2002

NetLeverage
NetLeverage

AD replication error 8606 "Insufficient attributes were given to create an object": find and remove lingering objects

Technical article
Troubleshooting

About this article: This article is reviewed against the cited sources. This is source research, not lab validation. Check the stated versions and your own requirements before making changes.

Error 8606 means a source domain controller tried to send a change to an object that the destination domain controller deleted long ago and has already purged. Because the destination runs in strict replication consistency, it refuses to bring the object back and stops replicating that partition from that source.

The most important point, and the one most often missed: the error is logged on the destination, but the object that needs removing is on the source (Microsoft Learn, error 8606). Cleaning the domain controller that shows the error achieves nothing.

What you see

  • dcdiag fails the Replications test with "The replication generated an error (8606): Insufficient attributes were given to create an object. This object may not exist because it may have been deleted and already garbage collected."
  • repadmin /showrepl, /replsum and /syncall report 8606, and Replicate now in Active Directory Sites and Services shows the same message.
  • Event 1988 (source NTDS Replication) on the destination: "The local domain controller has attempted to replicate the following object from the following source domain controller. This object is not present on the local domain controller because it may have been deleted and already garbage collected." It names the source domain controller, the object's distinguished name and its object GUID.

Users may also notice side effects of the lingering objects themselves: a deleted account still showing in the Exchange global address list, duplicate entries with CNF: in their names, mail not delivered to a renamed user, or a group that was deleted still granting access through users' tokens (Microsoft Learn, lingering objects).

Why it happens

A lingering object is an object that was deliberately deleted on at least one domain controller but still exists on another, because that domain controller never received the deletion within the tombstone lifetime. Microsoft names five ways to get one (Microsoft Learn, error 8606):

  1. Long replication failure or downtime. A source domain controller was offline or failing replication for longer than the tombstone lifetime, then came back and sent updates to objects the rest of the forest had purged. This is the most frequent cause.
  2. An object at the edge of the tombstone lifetime. Some operations, such as an in-place upgrade or a change to the partial attribute set, touch every object including recently deleted ones. Microsoft notes that adding the first domain controller of a new version to an existing forest is a known trigger. These lingering objects are transient and clear when the source next runs garbage collection.
  3. A clock jump on a domain controller that made it purge deleted objects early.
  4. A restore or undelete at the edge of the tombstone lifetime, for example an authoritative restore from a system state backup taken almost a tombstone lifetime ago.
  5. A USN bubble, where changes made during a rollback window never replicated out.

The 1988 event names only the object currently at the front of the queue. Microsoft warns there are "likely additional objects 'behind'" it, so expect more than one.

How to prove the cause

All of these checks are read-only.

1. Read the tombstone lifetime.

repadmin /showattr "CN=Directory Service,CN=Windows NT,CN=Services,CN=Configuration,DC=<forest root domain>,DC=<TLD>" /atts:tombstonelifetime

2. List the blocking objects. Filter the Directory Service log on each destination for event 1988. For every distinct object, record the object's distinguished name, its GUID, the source domain controller and the partition.

3. Read the object's history on the source.

repadmin /showobjmeta <source DC FQDN from the 1988 event> "<GUID=object GUID from the 1988 event>"

Two attributes tell you what happened (Microsoft Learn, error 8606):

  • LastKnownParent: its date stamp is when the object was deleted.
  • IsDeleted: its date stamp is when the object was last deleted or reanimated, and its version shows which. Version 1 is the first deletion; in Microsoft's worked example, version 2 is the restore that brought the object back. Read the version together with whether the object is currently live or deleted on the source.

Search by GUID rather than distinguished name. The object may have been renamed, moved, conflict-mangled with CNF: or sit in the Deleted Objects container on the source.

4. Classify each object.

What /showobjmeta and the logs show What it is Action
Live on the source, deleted on the destination longer ago than the tombstone lifetime Permanent lingering object Remove it (below)
In the source's Deleted Objects container, deleted right at the tombstone lifetime boundary Transient lingering object Wait for garbage collection, or trigger it
IsDeleted at version 2 (restored), dated about a tombstone lifetime after LastKnownParent Reanimated at the boundary Decide whether the object should exist at all
Date stamps in the future, before the OS release, or with unexplained gaps Clock jump Fix time first

5. Confirm the scope without changing anything.

repadmin /removelingeringobjects <DC holding the lingering objects> <DSA GUID of a DC with a writable, healthy copy> <partition DN> /advisory_mode

Advisory mode lists what would be removed.

The fix

Transient lingering objects

Garbage collection runs on each domain controller every 12 hours from its last start, so transient lingering objects clear on their own. To speed this up, Microsoft documents triggering it on the source:

repadmin /setattr "" "" doGarbageCollection add 1

Permanent lingering objects

Microsoft names three tools (Microsoft Learn, error 8606):

  • Lingering Object Liquidator v2 (LoLv2). Microsoft's preferred tool. It is a graphical tool, downloaded separately, that scans the forest, detects and removes lingering objects.
  • repadmin /removelingeringobjects, without /advisory_mode. The argument order trips people up. The first argument is the domain controller that holds the lingering objects (usually the source named in event 1988). The second is the DSA object GUID of a domain controller with a writable, correct copy of the partition. The domain controller being cleaned must be able to reach that reference domain controller directly on TCP 389.
  • repadmin /rehost for a global catalog's read-only copy of another domain's partition. It rebuilds that read-only copy from a good writable source. Microsoft notes lingering objects are found most often in read-only domain partitions on global catalogs.

Repeat on every domain controller that holds lingering objects, and for each partition, then confirm replication as below.

Objects reanimated at the boundary

If the object was deliberately deleted, delete it from the source domain controllers that still hold a live copy. If it should exist everywhere, Microsoft's options are to enable loose replication on the destinations that lack it, force-demote the domain controllers that have already purged it, or delete and re-create the object. Beware of reanimating security principals that were intentionally deleted.

How to confirm the fix

  • No new event 1988 entries on the destination after the next replication cycle.
  • repadmin /showrepl shows a fresh last success for the partition from that source; repadmin /replsum shows no 8606.
  • A second repadmin /removelingeringobjects … /advisory_mode (or a LoLv2 scan) finds nothing.
  • The symptoms users saw (ghost accounts in the address list, duplicate CNF: entries) are gone.

What to watch afterwards

  • Daily replication health. Microsoft's guidance is to put domain controllers that have not replicated in 50 per cent of the tombstone lifetime on a priority watch list, and to force-demote those that cannot be made to replicate by 90 per cent. Microsoft names repadmin /showrepl * /csv as an inexpensive but effective way to do it.
  • Strict replication consistency stays on. It is what turned a silent reappearance of deleted accounts into a visible error. Check it is still enabled after you finish.
  • Time configuration. If a clock jump caused it, follow Microsoft's checks: the forest root PDC emulator uses a reliable external time source, domain controllers take time from the domain hierarchy, and virtual machines follow the hypervisor vendor's time guidance.

Risks and limits

  • Never switch a destination to loose replication consistency just to clear 8606. That would bring the deleted objects back into the forest rather than remove them.
  • Pick the reference domain controller carefully. removelingeringobjects removes anything on the target that the reference does not have, so the reference must be the one you trust.
  • Microsoft also mentions repldiag as a way to automate removal across a forest, but advises keeping control of the process in larger environments.
  • If repadmin does not remove the object reported in event 1988, Microsoft points to two further checks: whether the object was created in a USN gap, and whether the object's originating domain controller is missing from the source's up-to-dateness vector. That is specialist territory.

Sources

Next step

Lingering objects are rarely the only problem: something stopped replication for months first. If you need the blocking objects identified, removed from the right domain controllers and replication monitored from now on, NetLeverage has looked after business servers since 2002. Talk to us through managed support with the text of one 1988 event and your repadmin /showrepl * /csv output.