Sql-server – SQL Server Deadlock on two updates due to index lock order

deadlocksql server

I have two UPDATEs – one locks the CI first and then the NCI (on status) because the status column is also being updated. The other already owns a U lock on the NCI because it knows it is changing and then tries to get a U lock on the CI.

What is the easiest way to force these to serialize? It seems odd to use a TABLE-level hint since this is an internal indexing issue – there is only one table involved – will UPDLOCK, HOLDLOCK automatically just apply to all indexes needed on that table and thereby force it to be serialized?

Here are the queries:

UPDATE htt_action_log
SET status = 'ABORTED', CLOSED = GETUTCDATE()
WHERE transition_uuid = '{F53ADDDA-E46B-4726-66D8-D7B640B66597}'
AND status = 'OPEN';

That one X locks the row in the CI (on CREATED column) and then attempts to X lock on the NCI which includes the status column.

UPDATE htt_action_log
SET status = 'RUNNING {36082BCD-EB52-4358-E3D3-4D96FD5B9F0F} 1360094342'
WHERE action_uuid = (SELECT TOP 1 action_uuid
                     FROM htt_action_log
                     WHERE transition_uuid = '{F53ADDDA-E46B-4726-66D8-D7B640B66597}'
                         AND status = 'OPEN'
                     ORDER BY action_seq)

This one U locks the same NCI – for the nested query I guess, then goes to lock the CI for the update.

Thus the order produces the deadlock.

Easiest solution is to force the two queries to completely block – i.e. serialize. What's the easiest way to force that, just put WITH (UPDLOCK, HOLDLOCK) on the references to the table (one in the first and two in the second)?

DDL:

Note client has more indexes on this table which should be affected by this update, but are not mentioned in the deadlock graph.

CREATE TABLE [dbo].[HTT_ACTION_LOG](
    [ACTION_UUID] [varchar](128) NOT NULL,
    [TRANSITION_UUID] [varchar](128) NOT NULL,
    [STATUS] [varchar](128) NOT NULL,
    [CREATED] [datetime] NOT NULL,
    [CLOSED] [datetime] NULL,
    [ACTION_SEQ] [int] NOT NULL,
    [ACTION_TYPE] [varchar](15) NOT NULL,
    [ACTION_NAME] [varchar](50) NOT NULL,
    [ACTION_RESULT] [varchar](8000) NULL,
    [PENDING_SINCE] [datetime] NULL,
    [ACTION_SQL] [varchar](8000) NULL,
    [ERROR_OK] [int] NULL,
    [ERROR_COND] [varchar](2048) NULL,
    [RETRY] [varchar](128) NULL,
 CONSTRAINT [PK_HTT_ACTION_LOG_1] UNIQUE NONCLUSTERED 
(
    [ACTION_UUID] ASC
)
)

CREATE CLUSTERED INDEX [IK_HTT_ACTION_LOG_2] ON [dbo].[HTT_ACTION_LOG] 
(
    [CREATED] ASC
)

CREATE NONCLUSTERED INDEX [IK_HTT_ACTION_LOG_1] ON [dbo].[HTT_ACTION_LOG] 
(
    [TRANSITION_UUID] ASC,
    [STATUS] ASC
)
INCLUDE ( [ACTION_UUID],
[ACTION_SEQ])

CREATE NONCLUSTERED INDEX [IK_HTT_ACTION_LOG_4] ON [dbo].[HTT_ACTION_LOG] 
(
    [ACTION_UUID] ASC,
    [STATUS] ASC
)

CREATE NONCLUSTERED INDEX [missing_index_11438530_11438529_HTT_ACTION_LOG] ON [dbo].[HTT_ACTION_LOG] 
(
    [TRANSITION_UUID] ASC,
    [ACTION_TYPE] ASC
)
INCLUDE ( [ACTION_NAME])

CREATE NONCLUSTERED INDEX [missing_index_7207590_7207589_HTT_ACTION_LOG] ON [dbo].[HTT_ACTION_LOG] 
(
    [STATUS] ASC
)
INCLUDE ( [CREATED],
[PENDING_SINCE],
[ACTION_NAME])

CREATE NONCLUSTERED INDEX [missing_index_8535421_8535420_HTT_ACTION_LOG] ON [dbo].[HTT_ACTION_LOG] 
(
    [TRANSITION_UUID] ASC
)
INCLUDE ( [ACTION_UUID],
[STATUS])

ALTER TABLE [dbo].[HTT_ACTION_LOG] SET (LOCK_ESCALATION = AUTO)

ALTER TABLE [dbo].[HTT_ACTION_LOG]  WITH CHECK ADD  CONSTRAINT [FK_HTT_ACTION_LOG_1] FOREIGN KEY([TRANSITION_UUID])
REFERENCES [dbo].[HTT_TRANSITION_LOG] ([TRANSITION_UUID])

ALTER TABLE [dbo].[HTT_ACTION_LOG] CHECK CONSTRAINT [FK_HTT_ACTION_LOG_1]

ALTER TABLE [dbo].[HTT_ACTION_LOG] ADD  DEFAULT ('OPEN') FOR [STATUS]

ALTER TABLE [dbo].[HTT_ACTION_LOG] ADD  DEFAULT (getutcdate()) FOR [CREATED]

ALTER TABLE [dbo].[HTT_ACTION_LOG] ADD  DEFAULT ((0)) FOR [ERROR_OK]

Best Answer

The optimal index for those two queries is not far from the existing definition of the IK_HTT_ACTION_LOG_1 index (add ACTION_UUID as an INCLUDE to the improved index below):

CREATE INDEX nc1
ON dbo.HTT_ACTION_LOG
(
    TRANSITION_UUID,
    STATUS,
    ACTION_SEQ
);

The first query is:

UPDATE dbo.HTT_ACTION_LOG
SET [STATUS] = 'ABORTED', 
    CLOSED = GETUTCDATE()
WHERE
    TRANSITION_UUID = '{F53ADDDA-E46B-4726-66D8-D7B640B66597}'
    AND [STATUS] = 'OPEN';

Giving the following execution plan:

Update 1

The second query can be expressed this way:

UPDATE ToUpdate 
SET [STATUS] = 'RUNNING {36082BCD-EB52-4358-E3D3-4D96FD5B9F0F} 1360094342'
FROM
(
    SELECT TOP (1)
        hal.[STATUS]
    FROM dbo.HTT_ACTION_LOG AS hal
    WHERE
        hal.transition_uuid = '{F53ADDDA-E46B-4726-66D8-D7B640B66597}'
        AND hal.[STATUS] = 'OPEN'
    ORDER BY
        hal.ACTION_SEQ ASC
) AS ToUpdate;

Giving this execution plan:

Update 2

Both queries now access the same resources in the same order, while locking many fewer rows on the read side of the plan. The execution engine will automatically take UPDLOCKs when reading the new index, providing the serialization you are looking for.

Related Question