[PATCH] Fix ldb lockout after DRS failures

[Andrew Bartlett]

On Sat, 2013-01-26 at 11:44 +1100, Andrew Bartlett wrote:
> This patch, required for my server_id patches (apparently) just passed
> an autoubuild.  It is designed to address this message Matthieu got when
> trying to autobuild my second server_id patch.
> 
> In short, as we fix the issues that were really breaking DRS, we are
> finding issues with running DRS.  It is a can of worms, but we are
> making progress, and if we can get these into 4.0.2, then we really will
> have a much more reliable replication system.
> 
> The issue was that if a single DRS replication failed at the pre-commit
> stage, that we would get into a limbo-land were one half of ldb thought
> there was still an active transaction, while another half thought there
> wasn't.  That ldb handle then became read-only until closed. 

This is the log that lead me to this error:

IRPC callback failed for DsReplicaSync - NT_STATUS_IO_TIMEOUT
../source4/dsdb/repl/replicated_objects.c:653 Failed to prepare commit of transaction: DRS linked attribute for GUID 81bb9d15-30f4-4653-9b4f-fc1e50cfbbf8 - DN not found
Failed to commit objects: WERR_GENERAL_FAILURE/NT_STATUS_INVALID_NETWORK_RESPONSE
../source4/dsdb/repl/drepl_ridalloc.c:43: RID Manager failed RID allocation - WERR_BAD_NET_RESP - extended_ret[0x1]
Failed to store repsTo - ltdb modify without transaction
Failed to delete repsTo for da6239d9-d141-4a05-a9e8-f10694d3a763: WERR_DS_DRA_INTERNAL_ERROR
UpdateRefs failed with WERR_DS_DRA_INTERNAL_ERROR/NT code 0xc00020fa for da6239d9-d141-4a05-a9e8-f10694d3a763._msdcs.samba.example.com DC=DomainDnsZones,DC=samba,DC=example,DC=com
Failed to apply records: partition_metadata: increment sequence number without transaction at ../source4/dsdb/samdb/ldb_modules/partition_metadata.c:346: Operations error
Failed to commit objects: WERR_GENERAL_FAILURE/NT_STATUS_INVALID_NETWORK_RESPONSE
Failed to store repsFrom - ltdb modify without transaction
Failed to apply records: ltdb modify without transaction: Operations error
Failed to commit objects: WERR_GENERAL_FAILURE/NT_STATUS_INVALID_NETWORK_RESPONSE
Failed to store repsFrom - ltdb modify without transaction
Failed to apply records: ltdb modify without transaction: Operations error
Failed to commit objects: WERR_GENERAL_FAILURE/NT_STATUS_INVALID_NETWORK_RESPONSE
Failed to store repsFrom - ltdb modify without transaction
Failed to store repsTo - ltdb modify without transaction
Failed to delete repsTo for da6239d9-d141-4a05-a9e8-f10694d3a763: WERR_DS_DRA_INTERNAL_ERROR
UpdateRefs failed with WERR_DS_DRA_INTERNAL_ERROR/NT code 0xc00020fa for da6239d9-d141-4a05-a9e8-f10694d3a763._msdcs.samba.example.com DC=samba,DC=example,DC=com
Failed to apply records: partition_metadata: increment sequence number without transaction at ../source4/dsdb/samdb/ldb_modules/partition_metadata.c:346: Operations error

Andrew Bartlett

-- 
Andrew Bartlett                                http://samba.org/~abartlet/
Authentication Developer, Samba Team           http://samba.org