An SQL text by Erland Sommarskog, SQL Server MVP. Latest revision: 2017-04-08.
If you ever tried to track down a blocking situation by running sp_who, DBCC INPUTBUFFER and the various DMVs while the users are screaming, beta_lockinfo is the thing for you.
Table of Contents
beta_lockinfo is a stored procedure that provides information about processes and the locks they hold as well their active transactions. beta_lockinfo is designed to gather as much information about a blocking situation as possible, so that you can instantly find the culprit and kill the blocking process if the situation is desperate. Then you can sit back and analyse the output from beta_lockinfo to understand how the blocking situation arose and figure out what actions to take to prevent the situation from reoccurring. The output from beta_lockinfo shows all active process as well as passive processes with locks, which objects they lock, what command they last submitted and which statement they are executing. You also get the query plans for the current statements. Normally, you run beta_lockinfo to look at the output directly, but there is also an archive mode where the data is saved to table. This is not the least useful, if you want someone to send you the output from beta_lockinfo at a site you don't have access to yourself.
The latest version is SourceSafe version 22, checked in 2017-04-08 19:02.
There are four versions of beta_lockinfo. Download the one that fits your SQL Server version:
To export the data from beta_lockinfo to a file that you can mail, use any of the BAT files included in beta_lockinfo.zip. See the section Using Archive Mode for more details.
For older versions of SQL Server (down to SQL 6.5), use my older aba_lockinfo instead.
You can put the procedure in any database you like as long as the database has compatibility level 90 or higher.
beta_lockinfo compiles information from a number of DMVs in this order:
Note that since beta_lockinfo reads the information at different points in time, the various pieces will not be fully consistent. For instance, a new process may log in and grab a lock after beta_lockinfo has read the process information but before beta_lockinfo reads the locks. Or the statement that was current when the process information was collected may not match the locks you see, because the process has moved on.
For the impatient, here is a complete list of the parameters. And if these five are too many for you, don't be worried. You can run beta_lockinfo without any parameters, and this is what I do myself all the time.
|If 0 (the default) beta_lockinfo, lists only "interesting" processes. If 1, beta_lockinfo also lists "half-interesting" processes. If ≥ 2, beta_lockinfo lists all processes. See the section below for the definition of "interesting" and "half-interesting" processes.|
|If 1, the result set is returned to make it easier to read when you have set SSMS to return data in text mode: columns are trimmed to be as wide as the widest data in the column (an idea taken from sp_who2), and there is a blank line inserted between each process. If 0 (the default), the result set is returned without these extra thrills. This mode is also known as grid mode, since in you would typically use this setting when you have selected grid mode in your query tool.|
When this column has a non-NULL value, beta_lockinfo runs in archive mode. No data is returned as you run the procedure. Instead the data is saved to the table guest.beta_lockinfo. (The guest schema exists in all databases, and cannot be dropped.) This permits you take several consecutive snapshots for later analysis. It is also useful if you want someone at a remote site to send you data for analysis. The value of the parameter specifies that data in the table that is older than @archivemode minutes should be deleted from the table. A special case is when you specify 0: in this case beta_lockinfo does not collect any data, but drops and recreates the table. For more details on archive mode, see the section Using Archive Mode.
Archive mode is incompatible with @textmode = 1.
|If, 1 beta_lockinfo prints progress information about its operations. If you find that beta_lockinfo takes a long time to run, I'm interested to see the output when @debug = 1, particularly if any other step than gathering lock information needs a lot of time.|
An interesting process is a process which is doing something that potentially could affect other processes directly. That is, a process that fulfils any of these conditions:
To clarify the two exception points for an active task: normally you don't want to see the process that runs beta_lockinfo in the output. That is, as long as you are not disrupting anyone else. If you previously had started a transaction and had acquired a lock, you will see yourself listed. (If someone else is running beta_lockinfo in parallel with you, you will see that process.)
The wait type BROKER_RECEIVE_WAITFOR is the typical state for a Service Broker procedure that is waiting for a message to arrive. In an application using Service Broker, you can have quite a few of these, and including them in the default output has proven to produce too much noise.
Why would a process be missing from sys.dm_exec_sessions? There are a couple of reasons why this could happen.
A half-interesting process is a process which has some level of activity, but is not likely to directly affect other processes. Any of the following conditions qualifies a process to be added to the output when @allprocesses = 1:
Up to SQL 6.5, life was very easy. A process was a connection was a session. These days there are a whole bunch of different concepts: sessions, execution contexts, requests, connections and tasks. At least. This section describes how beta_lockinfo uses these concepts.
Let's start with a connection, because it's the easiest in this context: beta_lockinfo completely ignores the DMV sys.dm_exec_connections, right or wrong.
The main concept is the session, as listed in sys.dm_exec_sessions. A session is identified with a session_id, or what old-time users of SQL Server knows as a spid.
A session may be idle, or it may be executing one or more requests. A request is a batch of SQL commands or an RPC call to a stored procedure. In the very most cases, a session runs only one at request at a time, but SQL 2005 introduced MARS, Multiple Active Result Sets. When MARS is in effect, a session may submit a new command batch, even if the previous batch has not completed. The batches are executed in an interleaved fashion. Requests appear in the view sys.dm_exec_requests. A request is identified by the columns session_id and request_id. As long as MARS is not in effect, request_id is usually 0. When MARS is in effect the requests are numbered 1, 2 etc. Beside MARS, I have noticed one more case when request_id is non-zero: requests executed by sessions connected through HTTP endpoints. (A feature introduced in SQL 2005, deprecated in SQL 2008 and removed in SQL 2012.)
An SQL command can be executed single-threaded or by parallel threads. Each such thread is known as an execution context, identified by the exec_context_id. The main thread for the request has exec_context_id = 0, and any non-zero number for an exec_context_id indicates parallel execution. There is no view as such that defines the execution contexts, but all execution contexts have a row in sys.dm_os_tasks. Now, here is a funny thing: if a request with a non-zero request_id requires a parallel plan, the request_id for the threads with a non-zero exec_context_id is nevertheless zero, meaning that if a session has multiple requests running, you cannot tell which request the non-zero execution contexts belong to. (I assume, though, that all contexts belong to the same request, since multiple requests are interleaved, and not truly multi-process.)
A task is something that is listed in sys.dm_os_tasks, and is identified with a varbinary(8) value. All execution contexts, and thus all requests, have a row in sys.dm_os_tasks. There are also non-session entries in sys.dm_os_tasks. beta_lockinfo displays these if they are waiting and the wait type is an interesting wait type. That is, it is not a wait type commonly used by system processes that are almost always waiting.
In the output, beta_lockinfo includes session_id, exec_context_id and request_id. But instead of one column for each, beta_lockinfo merges them into one string with the ids separated by slashes, for instance 56/0/2 (request 2 for session 56, the main thread) or 65/7/0 (execution context 7 for session 65, unknown for which request). In the very common case that the both exec_context_id and the request_id are 0, beta_lockinfo suppresses them, and lists only the session_id. Thus, you will not see 64/0/0, only 64. Session-less tasks are displayed with a session id that is < -1000. You may also see -9 for exec_context_id. This happens when there is a session-less task in sys.dm_os_waiting_tasks that can be connected back to a session through the column parent_task_address in sys.dm_os_tasks. (This never occurs on SQL 2005 as parent_task_address is not available in this version.)
In the output for a process, any non-zero requests appear first, with one line per request. Next follow one or more lines with the session_id proper. There is always at least one such line for a session, even if there is no row for the session in sys.dm_os_tasks. Finally, any execution contexts with a non-zero exec_context_id for the session are listed in order. Occasionally, you may see the same exec_context_id listed twice. This is because two tasks can have the same exec_context_id. (Why it is so, I don't know. Maybe it's a bug. But I've seen it happen.)
Locks are displayed per session. In sys.dm_tran_locks they are listed per execution context and request, but I've opted to collapse them on session level to reduce the number of lines in the output for a massively parallel query. Since it's not possible to (easily) determine to which request an execution context belongs to, it was logical to collapse the requests as well. After all, I don't expect multiple requests to be that common. (Then again, that may reflect my attitude against MARS, a feature that I find to be of dubious value.)
Finally, there is process, which is not really an SQL Server concept, but which I use in this document somewhat loosely. Sometimes I refer to entire session, sometimes to the unique combination of session, execution context and request.
The key column in the result set is the blklvl column. If this column has
double exclamation marks
!! this process is a lead blocker, blocking
one or more other processes (which could be other execution contexts of the same
session) without being blocked itself. If there is
a number, the process is blocked. 1 indicates that the process is blocked by a
lead blocker, 2 indicates that the process is blocked by a process which in its
turn is blocked by a lead blocker and so on. A process is not involved in
blocking has a blank in this column.
The values in blklvl may be in parentheses. That is, instead of
(!!). This happens when you have a parallel query, where
the various threads are blocking each other,
but they are neither blocked by any other session, nor blocking any other session.
Thus, if you are investigating a blocking issue, you can ignore sessions where
blklvl values are in parentheses.
The column blkby shows which process the process is directly blocked by. The blkby column is formed as a string with session_id, exec_context_id and request_id concatenated with slashes, and with the latter two suppressed if both are zero. Sometimes you may see a string like 54/-1/-1. This means that when beta_lockinfo read sys.dm_os_waiting_tasks, there was a blocking task, but when beta_lockinfo read sys.dm_exec_sessions, sys.dm_os_tasks etc, the task had exited. You may also see something like (+4) after the process string. This means that the process is in fact blocked by four more processes but the one listed. (Which one that is listed when there are several blockers is fairly arbitrary.)
There are two other possible values for blklvl: DD and ??. DD means that the process is involved in an on-going deadlock. If you look at blkby, and then go to that spid, look at its blkby and so on, you can expect to come back where you started sooner or later. If a process is blocked by a deadlocked process without being involved in the deadlock itself, blklvl displays 1, 2 etc as with regular blocking. The value ?? means that the process was listed in sys.dm_tran_locks as waiting for a lock, but it was not listed in sys.dm_os_waiting_tasks. This indicates that by the time beta_lockinfo read sys.dm_os_waiting_tasks, the process had not yet requested the lock.
To see what a process is waiting for, look at the lstatus and waittype columns. In most cases, a process is waiting for a lock, which is indicated by WAIT or CONVERT in the lstatus column. All non-granted locks appear first in the output for a process. In the case of parallel execution, the main thread or a sub-thread may be waiting for another thread to complete. In this case waittype will be CXPACKET.
As outlined above, the result is organised so that there is essentially one row per group of locks. Locks are aggregated on the following columns in sys.dm_tran_locks: resource_type, resource_subtype,resource_database_id, resource_description (application locks only) resource_associated_entity_id (for the resource types it applies to), request_mode, request_status and request_owner_type.
To this come rows for processes with non-zero request ids, listed before the locks, and processes with non-zero execution context ids, listed after the locks. When the parameter @procdata is 'F', process-data is listed only for the first lock row, as well for any request rows listed before the locks. But it is not repeated for every subsequent lock row or rows for execution contexts.
In several cases, NULLs and other values that mean "nothing" are displayed with blank space to make the output easier to read. I've retained NULL in the cases where I think NULL has something to tell.
|spid||The process id with the execution-context id and the request id added, if any of them are non-zero, as described above. Numbers < -1000 are spid-less tasks that are waiting for something.|
|command||The current command for the request, taken from sys.dm_exec_requests.request_command. This is the same as you see in sp_who, except that for an idle process, you will not see AWAITING COMMAND, but a NULL value.|
The login the process logged into to SQL Server as. If the process has engaged in impersonation through EXECUTE AS, either through an explicit statement or by calling a stored procedure with an EXECUTE AS clause, the currently impersonated login is shown in parentheses. If the impersonation is on user level, you will see a SID in the parentheses.
If the process is a system process (sys.dm_exec_sessions.is_user_process = 0), login reads SYSTEM PROCESS, and any impersonated login is not shown.
|host||The computer the session connected from. Keep in mind that most APIs permits the application/user to set the host name to whatever they like.|
|hostprc||The PID (process id) of the client application. Taken from sys.dm_exec_sessions.host_process_id.|
|endpoint||The endpoint for the process. Most of the time you will see things like TSQL Default TCP. Taken from sys.endpoints.name via sys.dm_exec_sessions.endpoint_id.|
|appl||The name the client application used to identify itself when it connected.|
|dbname||The current database for the process. If there are requests with non-zero ids, it can happen that the request has a different current database than the main session. The database for a running request is taken from sys.dm_exec_requests.database_id. The session-level database is taken from sys.dm_exec_sessions.database_id. (sys.sysprocesses.dbid on SQL 2005/2008.)|
|prcstatus||The status of the process. For a running request, this is taken from sys.dm_os_tasks.task_state. For an idle session without a request, the data is taken from sys.dm_exec_sessions.status. You can easily tell them apart, as former is all uppercase and the latter is all lowercase.|
|ansiopts||A list of ANSI-related SET options that deviate from the default setting. These are the values you could encounter:
|spid_||This is the spid column repeated for convenience. However, there is one case where this column will be NULL, and that is when the session appears in sys.dm_tran_locks only. In this case, only the first spid column has the value.|
Lists transaction-related SET options with values that deviate from the default settings. There are three fields. The first of them reflects SET TRANSACTION ISOLATION LEVEL. The value can be any of Read uncommitted, Repeatable read, Serializable, Snapshot and Unspecified. (As for the latter value, I don't know what it is, but it's listed in Books Online) When the isolation level is the default Read Committed, nothing is listed.
The second field reflects SET LOCK_TIMEOUT, and displays the lock timeout prefixed with LT=. If no lock timeout is set (= -1) this is not displayed.
The third field is for SET DEADLOCK_PRIORITY and displays the chosen priority prefixed by DP=. The default priority of 0 is not displayed.
|opntrn||The value of @@trancount for the process, blank if this is 0. Note that if a process has a value here, but has NULL in command, and thus is idle, this indicates an orphaned transaction, something which can lead to trouble if the process is holding locks.|
|trninfo||A hyphenated value of five fields giving various titbits about the transaction. Please see the section Transaction Information below.|
|blklvl||When blank the process is
not blocked and is not blocking anyone. Double exclamation marks |
|blkby||Which process the process is blocked by. A string that is formed from the session id, execution-context id and the request id in the same manner as spid.|
|cnt||The columns following cnt detail a group of locks, and the cnt column is the number of lock in such a group.|
The object on which the process has a lock, or is waiting to get a lock on. In most cases, you will see database.schema.table.index here. If the lock is on a key in a clustered index, beta_lockinfo suppresses the index name, to remind you that a key lock in a clustered index is really a row lock on the data.
If you see a number in parentheses, the table is partitioned, and the number is the partition number of the table or index. Even if the table/index is partitioned, beta_lockinfo never shows partition number 1.
If the lock is on an allocation unit, beta_lockinfo includes sys.allocation_units.type_desc in brackets in the string.
For some resource types, you will only see the database name, because the lock is on database level, or on an object which is not schema-bound, for instance an XML schema collection. You may also only see the database name if the entity id in sys.dm_tran_locks did not translate, for instance because the object disappeared while beta_lockinfo was running.
Temp tables are displayed without any database prefix, since they always are in tempdb. If you see something like #temp (x3), this means that the process has created three different temp tables with the same name. This can happen if you have a recursive procedure, but also if a process inside a transaction repeatedly calls another procedure that creates a temp table. The three temp tables are distinct entries in tempdb.sys.objects, but beta_lockinfo aggregates the entries to reduce the number of rows in the output.You may also see things like #temp[PK] (x20). This refers the metadata lock held on the primary key for the twenty temp tables. Rather than PK you may also see U, C and D for Unique, Check and Default constraints.
Table variables are entered in the system catalogue as # + eight hex digits. This combination can also refer a temp table that has been dropped, but where SQL Server has cached the table definition. beta_lockinfo aggregates all these tables to a single row per type of lock with the text #(tblvar or dropped temp table) followed by, for instance (x4), if there is more than one such table.
Sometimes you may see things like dbname.<12345978>. The number in that case is an object id that beta_lockinfo was not able to find. This could be because the table was dropped before the beta_lockinfo came round to translate the names. It can also occur if you run beta_lockinfo with reduced permissions, see below.
For application locks, the value is taken straight from sys.dm_tran_locks.resource_description.
An object may have been created inside an on-going transaction, and the metadata about the table is still locked. beta_lockinfo uses READ UNCOMMITTED to access the metadata, but Books Online makes it clear there are no guarantees that any other isolation level than READ COMMITTED is supported for access to the catalog views. Nevertheless, it seems to work without problems up to SQL 2012. On SQL 2014 and later it is not possible to get information about an uncommitted heap from sys.partitions. In this case, beta_lockinfo will display <heap in trans> prefixed by # for tempdb and the database name for other databases.
If an error occurs when retrieving the object name, you will see an error message in this column.
|rsctype||What sort of resource that is locked. Taken from sys.dm_tran_locks.resource_type. See the topic for sys.dm_tran_locks in Books Online for details.|
|locktype||What sort of lock the process has taken out/is waiting to get. Taken from sys.dm_tran_locks.request_mode. See the topic Lock Modes in Books Online for details.|
|lstatus||Whether the lock has been granted or not. Taken from sys.dm_tran_locks.request_status. In order to make WAIT more visible, beta_lockinfo puts the very common grant in lowercase.|
|ownertype||The entity that owns the lock. Taken from sys.dm_tran_locks.request_owner_type, except that beta_lockinfo shortens SHARED_TRANSACTION_WORKSPACE to STW, as almost all processes have a Sch-S on database level with this owner type.|
|rscsubtype||This column holds two pieces of locking information:
|waittime||How long the process/request has been waiting in seconds (with three decimals). Blank if the process is not waiting. Taken originally from sys.dm_os_waiting_tasks.wait_duration_ms.|
|waittype||What the process/request is waiting for. Blank if the process is not waiting. Taken from sys.dm_os_waiting_tasks.wait_type.|
This column is only included on SQL 2016 and later. It holds the top 5 wait types for the session and the total wait time in ms for these wais.
Please take this information with a grain of salt. If the process has been connected for a long time, the waits may relate to past troubles and not be relevant to the current situation. Also, it is unclear whether all waits that occur during the running of a request are written back to sys.dm_exec_session_wait_stats from where the information is taken.
|spid__||Another repetition of the process id. NULL if the process appears in sys.dm_tran_locks only.|
|cpu||Taken from sys.dm_exec_sessions.cpu_time. If the process is running a request, beta_lockinfo shows the value of sys.dm_exec_requests.cpu_time in parentheses.|
|physio||The sum of the column reads and writes in sys.dm_exec_sessions. If the process is running a request, beta_lockinfo also displays the sum of reads and writes from sys.dm_exec_requests in parentheses.|
|logreads||Taken from sys.dm_exec_sessions.logical_reads. If the process is running a request, beta_lockinfo shows the value of sys.dm_exec_requests.logical_reads in parentheses.|
The memory grant for the running request in megabytes with three decimals, taken from sys.dm_requests.granted_query_memory. Blank for sessions that are not running requests.
|progress||How much of the command that has been completed in per cent. The value comes from sys.dm_exec_requests.percent_complete. SQL Server only presents a value for some commands like DBCC, BACKUP, RESTORE and ROLLBACK. See Books Online for a full list. Blank when the source column is 0.|
Space usage in tempdb. The column is the sum of the counter values in sys.dm_db_session_space_usage. More precisely that means the sum of allocated pages minus the sum of deallocated pages. The value can be negative in some situations. The value can also be inflated, because when a process fills a temp table with many rows and then drops it, the deallocations are credited to a system process that performs deferred space deallocation. (Sebastian Meine discusses this in detail in this excellent blog post.) If the process is running a request, beta_lockinfo shows the same sum from sys.dm_db_task_space_usage, aggregated per request in parentheses. (The DMV has value per task, but I opted to have a consistent display with other resource columns.)
|now||The value of sysdatetime() when beta_lockinfo started running, so you can relate to the other time columns. In grid mode and text mode, the value includes only the time portion. In archive mode, you get the full value.|
|login_time||When the processes originally logged into SQL Server. Taken from sys.dm_exec_sessions.login_time. The date is only displayed, if the login was on an earlier day. The format for the date is YYMMDD.|
|last_batch||When the process most recently submitted a command batch. Taken from sys.dm_exec_sessions.last_request_start_time. The date portion is handled as for login_time.|
|trn_start||If the process has an active transaction, this is when this transaction started. Taken from sys.dm_tran_active_transactions.transaction_begin_time.|
The difference between now and last_batch displayed as days + hh:mm:ss.fff with leading zeroes (and associated delimiters) stripped. This value can be negative, if the process submitted a batch after beta_lockinfo started executing but before it arrived at reading sys.dm_exec_sessions.
|trn_since||The difference between now and trn_start in seconds displayed in the same manner as last_since. This value can be negative, if the process started a transaction, after beta_lockinfo started executing.|
|clr||This column reads CLR, if the process has a CLR object (procedure, function, trigger etc) somewhere on the call stack. Or at least the text for the column sys.dm_exec_requests.executing_managed_code seems to say so. My experience does not really agree with Books Online. Rather the flag appears to be set if the request at some point has executed CLR code.|
|nstlvl||The current value of @@nestlevel for the process, if it is running a request.|
|spid___||Yet a repetition of the process id. NULL if the process appears in sys.dm_tran_locks only.|
|inputbuffer||The last command sent from the client, taken from sys.dm_exec_input_buffer (SQL 2016 and later) or DBCC INPUTBUFFER (earlier versions). In text mode, beta_lockinfo replaces newlines with spaces to keep the output on one line. Note that since beta_lockinfo runs this command after having collected the locks (which takes some time if there are many of them), the command you see here may not match other information in the output.|
The stored procedure, function or trigger that the process was executing when beta_lockinfo collected process information. This can be the same the procedure that you see in the inputbuffer column, but in many cases it is different due to nesting.
The column is blank if the process is not active, or if the current scope is a loose batch of SQL statements. Note that the latter includes dynamic SQL invoked by EXEC() or sp_executesql!
The statement that the process was executing when beta_lockinfo collected process information. If current_sp is non-blank, this is a statement in current_sp. If current_sp was created WITH ENCRYPTION, beta_lockinfo cannot retrieve the query text, but instead displays the text -- ENCRYPTED, pos start/stop, where start and stop are the positions in characters for where the statement starts and stops. If you have access to the source code, you can use these values to determine the statement.
The statement is displayed in full on the first row for the process, but for remaining occurrences only the first 50 characters are shown.
If curstmt is blank, this could be because the process is not active. But it could also be involved in activity outside a statement, for instance compilation of a stored procedure not in cache.
In text mode, beta_lockinfo replaces newlines with spaces to keep the output on a single line.
|queryplan||The query plan for curstmt as an XML document. This column is presented in grid and archive mode, but is absent in text mode. See the section Query Plans below for details.|
|rowno||This column appears only in archive mode. This is a numbering to retain the ordering used for grid mode and text mode.|
|In text mode there is a final blank column which is used to generate the blank lines to keep different processes apart.|
The data is sorted by spid (with non-zero request ids first, non-zero execution-context ids last), lstatus (with CONVERT and WAIT first), object and rsctype.
In SSMS, you can double-click the XML document and you will see the graphic plan. Beware though that it typically does not work if your version of SSMS is below the version of SQL Server you are connected to. Also, this feature was also non-functional with some versions of SSMS 2008 R2, See this blog post from SQL Server MVP Aaron Bertrand how to repair this. This feature is not available in SSMS 2005, but you will have to save the XML document with the extension .sqlplan and reopen it. Or install a newer version of SSMS...
If an error occurs when retrieving the plan, you will see a document with a single <ERROR> element. This can be due to locking, but also because the nesting level in the XML for the plan exceeds what is supported by the xml data type in SQL Server. (Interesting enough, the new sys.dm_exec_query_statistics_xml appears to be able to return such plans in the xml data type nevertheless.)
If you are on SQL 2014 or earlier, it is simple: the column queryplan will always show you the estimated query plan. While it can be helpful at times to reveal what is going on, it can also be deceivable, because the estimates may be wrong, and the plan is not as peaceful as it may seem.
SQL 2016 SP1 introduced a new DMV sys.dm_exec_query_statistics_xml which permits you to get actual values from a running query. This can help you to identify plans with bad estimates without having to wait for the query to complete. The actual values you see will be lower than the final actual, but in case of a failed estimate, the actual values may still exceed the estimates by a wide margin. Note however, that generation of actual values is not turned on by default. When collecting query plans, beta_lockinfo first attempts sys.dm_exec_query_statistics_xml and if it returns nothing, it gets the traditional estimated plan with sys.dm_exec_text_query_plan.
There are two different infrastructures in SQL Server to collect actual plans: the legacy infrastructure and the new lightweight infrastructure that was introduced simultaneously with sys.dm_exec_query_statistics_xml. There is a huge difference in impact of these. According to Microsoft, the new lightweight structure shows an overhead of merely 1.5 to 2 % for a TPC-C like workload, whereas the other one incurs an overhead of 90 %. Bear in mind that the exact overhead depends on the workload. The cost is higher for a workload with many short-running queries.
The new lightweight infrastructre can be activated in two ways:
The old legacy infrastructure can be activated in a number of ways:
It goes without saying that the lightweight infrastructure is to prefer. The SET commands make sense if you only want plans for a single process, but stay away from the other two legacy possibilities
Finally, keep in mind that what applies is when the query started running. That is, if you have a query that has been running for an hour and you want to look at the actual values, it will not help to turn on trace flag 7412 at this point.
The column trninfo holds information from various columns in the transaction DMVs that I have preferred to collapse into a single column to make it easier to read. The column can have up to five fields separated by hyphens:
|Transaction type|| This field reflects sys.dm_tran_active_transactions.transaction_type.
|Transaction state||Taken from sys.dm_tran_active_transactions.transaction_state. Books Online indicates that the possible values are 0 through 8, but I've seen other values when working with distributed transactions. In most cases you will see 2 (= active transaction) here. See the topic for sys.dm_tran_active_transactions in Books Online for a complete listing.|
|DTC state||Taken from sys.dm_tran_active_transactions.dtc_state. This field is only included if this column has a value <> 0, in which case the fields reads one of DTC:ACTIVE, DTC:PREPARED, DTC:COMMITTED, DTC:ABORTED, DTC:RECOVERD. This field indicates that the transaction is distributed, but interesting enough, the transaction type can be DIST, while the dtc_state column still is 0.|
|Bound||If the transaction is a bound transaction (that is, it shares the transaction space with another session), you see BND in the fifth field. (Or more probably the fourth, since I would not expect a bound transaction to have a value in dtc_state.)|
When using MARS, a session may have multiple active transactions. In that case, beta_lockinfo displays information about the oldest transaction.
In archive mode, beta_lockinfo writes the output to the table guest.beta_lockinfo in the same database as you have put beta_lockinfo. (The guest schema exists in every database and cannot be dropped.) This table has the same columns as the output in grid mode and the same formatting with two exceptions:
These two columns constitute the primary key of the table.
All columns but three in guest.beta_lockinfo are varchar or nvarchar with the collation Latin1_General_BIN2. The exceptions are the columns now (datetime2(3) on SQL 2008 and later, datetime on SQL 2005), queryplan (xml) and rowno (int).
When you run in archive mode, beta_lockinfo adds a final row with asterisks in all string columns, so that when you look at the output later, you can easily tell different executions apart. The row with asterisks is added even if beta_lockinfo collected no other rows to present. The columns now and rowno are populated for the asterisk rows like they are for other rows.
One way to use archive mode is to run beta_lockinfo several times in quick succession, for instance during a test, to look at the output later. You may also be tempted to run beta_lockinfo as a scheduled job or similar. But since beta_lockinfo formats the output for reading, the content of the table is not really optimal for analysis through queries. If you decide to schedule beta_lockinfo, I recommend that you don't run it more frequently than every five minutes. (See the section Permissions and Resources below for a discussion on how much resources beta_lockinfo takes.)
Another way to use archive mode is when you need to get information from a site that you don't have access to yourself. In this case you could ask the local DBA to run beta_lockinfo in archive mode, use BCP to copy out the data in the table and send the BCP file to you. This is better than just copying the output from grid mode into an Excel sheet, where the line breaks in the SQL statements cause a mess. To facilitate the use of archive mode at a site where the DBA is not so savvy, there is a zip file beta_lockinfo.zip, which includes:
The four beta_lockinfo.bat files all take four parameters of which the first is mandatory:
The DBA can run the BAT file several times, and each BCP operation will export the entire table. Note that if you want the DBA on the other site to collect data for more than 60 minutes, you need to instruct him or her to change the parameter value for @archivemode; recall that 60 means that data that is older than 60 minutes should be deleted from the table.
Generally, when you have received the BCP file, the simplest is to load it on a server running the same version of SQL Server as the source system, and then run:
EXEC beta_lockinfo @archivemode = 0 BULK INSERT guest.beta_lockinfo FROM 'C:\temp\beta_lockinfo.bcp' WITH (FORMATFILE = 'C:\temp\beta_lockinfo.fmt') -- Beware the version! SELECT * FROM guest.beta_lockinfo ORDER BY now, rowno
The first line drops guest.beta_lockinfo and recreates it without collecting any information from the server you are running it on. The BULK INSERT statement loads the data into the table. Be careful to change the statement to use the actual paths for your system. And be careful to use the matching format file.
It may not always be possible to find an SQL Server instance that matches the source system. In practice, you can load files from SQL 2005, SQL 2008, 2012 and SQL 2014 on any of SQL 2008, SQL 2012 and SQL 2014 using the script above. The tricky part is if SQL 2016 is in the mix, because the definition of guest.beta_lockinfo is different on this version (the column top5waits is extra, and the column inputbuffer has a different definition.) You can load a file created on SQL 2016 on an earlier version, down to SQL 2008, but you will need to create the table by culling the CREATE TABLE statement from the SQL 2016 version of beta_lockinfo. The same applies if you want to load files from earlier versions on SQL 2016. On SQL 2005, you can only load files created on SQL 2005. Finally, keep in mind that if you want to view query plans, it certainly helps if your version of SSMS is at least of the same version as the source system.
If you get a run-time error when running archive mode and you recently have installed a new version of beta_lockinfo, try dropping guest.beta_lockinfo as the table definition may have changed.
Normally, you would run beta_lockinfo from an account that is a member of the sysadmin fixed role or that has been granted CONTROL SERVER. However, you still can run beta_lockinfo from a less privileged account, as long you have the permission VIEW SERVER STATE. In case you are not running beta_lockinfo as sysadmin, you should beware of this:
Unfortunately, beta_lockinfo is not as lean as I would like it to be. The chief reason for this is that access to sys.dm_tran_locks is nowhere cheap when there are plenty of locks on the system. A quick test on my desktop, beta_lockinfo needs 32 seconds to complete on SQL 2008 when I have 2.3 million locks on an idle server (in a real case when you have that many locks, your server is anything but idle). The good news is that in SQL 2012 Microsoft reimplemented sys.dm_tran_locks to be faster, and for the same situation, beta_lockinfo completes in 12 seconds. Which still a long time when you have urgent situation to sort out.
There are some other queries that are potentially resource-intensive, since beta_lockinfo works hard on things that normally belong in the presentation layer. When I test on an idle server, all that is dwarfed compared to collecting the locking information. But on a server that is loaded, and particularly one where the CPU utilisation is 100 %, executing beta_lockinfo certainly will add even more pain to your server. To mitigate this, I have taken steps to make sure that no statement in beta_lockinfo will use a parallel plan, and I've also added hints to prevent recompilation.
If your system is really hogged, and you just want to know who is creating all this mess, try this:
SELECT request_session_id, COUNT(*) FROM sys.dm_tran_locks GROUP BY request_session_id
This completes in 10 seconds for the same situation on SQL 2008 on my hardware; on SQL 2012 it takes 8 seconds.
If you have questions or comments and not the least suggestions for improvements just mail me, Erland Sommarskog, on firstname.lastname@example.org. If you run into some error message or incorrect results, please understand that it may be difficult for me to reproduce the problem, since it is may be dependent on situations on your site that I was not able to predict. The more information you include, the better the chance that I will understand what happened. The output from @debug = 1 can be helpful.
IMPORTANT! If you have used Archive Mode, you must drop the table guest.beta_lockinfo, or else you will get the error String or binary data would be truncated. Note also that the format files have been updated.
Added a caveat to the description of the tempdb column. The stored procedure itself is unchanged.
Back to my home page.