FIX: You receive an "Error: 17883" error message when the checkpoint process delays the SQL Server database activity and does not yield the scheduler correctly (815056)

SYMPTOMS

The checkpoint process forces SQL Server to write all the dirty pages to disk. Systems that can support and maintain fast I/O rates, and that do not result in pending I/O attempts, can trigger a condition whereby the checkpoint process does not have to delay to wait for I/O processes to complete. When the checkpoint does not correctly yield, it can affect overall database activity and performance. This may result in reduced transaction and batch submittal rates. The following error message may appear in the SQL Server error log:

Error: 17883, Severity: 1, State: 0
The Scheduler 0 appears to be hung. SPID 7, ECID 0, UMS Context 0x045530B8.

Note The format of the 17833 message is different for separate versions of the SQL Server product; however, the meaning is the same.

This article discusses one possible cause of this error message. The fix that is described in this article only helps if you are encountering the same scenario.

RESOLUTION

Service pack information

To resolve this problem, obtain the latest service pack for Microsoft SQL Server 2000. For additional information, click the following article number to view the article in the Microsoft Knowledge Base:

290211 How to obtain the latest SQL Server 2000 service pack

Security patch information

This fix has been included with the security patch for Microsoft Security Bulletin MS03-031. For additional information about how to obtain that security patch from the Microsoft Download Center, click the following article numbers to view the articles in the Microsoft Knowledge Base:

815495 MS03-031: Cumulative Security Patch for SQL Server

821277 MS03-031:Security Patch for SQL Server 2000 Service Pack 3

Note Make sure that you read the security bulletin articles completely and thoroughly before you apply the security patch.

Hotfix information

The English version of this hotfix has the file attributes (or later file attributes) that are listed in the following table. The dates and times for these files are listed in coordinated universal time (UTC). When you view the file information, it is converted to local time. To find the difference between UTC and local time, use the Time Zone tab in the Date and Time tool in Control Panel.

CORE BUILD 8.00.780

   Date         Time   Version         Size             File name
   ----------------------------------------------------------------------------

   26-Feb-2003  03:41  2000.80.777.0      29,244 bytes  Dbmslpcn.dll
   08-Feb-2003  05:43                    786,432 bytes  Distmdl.ldf
   08-Feb-2003  05:43                  2,359,296 bytes  Distmdl.mdf
   30-Jan-2003  01:55                        180 bytes  Drop_repl_hotfix.sql
   01-Feb-2003  00:32  2000.80.763.0   1,557,052 bytes  Dtsui.dll
   30-Jan-2003  05:18                    746,470 bytes  Instdist.sql
   30-Jan-2003  01:55                      1,402 bytes  Inst_repl_hotfix.sql
   08-Feb-2003  06:40  2000.80.765.0      90,692 bytes  Msgprox.dll
   08-Feb-2003  03:53                  1,065,895 bytes  Replmerg.sql
   08-Feb-2003  06:40  2000.80.765.0     221,768 bytes  Replprov.dll     
   08-Feb-2003  06:40  2000.80.765.0     307,784 bytes  Replrec.dll
   30-Jan-2003  05:18                  1,084,318 bytes  Replsys.sql
   28-Feb-2003  01:34  2000.80.778.0     176,696 bytes  Sqlmap70.dll
   08-Feb-2003  06:40  2000.80.765.0      57,920 bytes  Sqlrepss.dll
   03-Mar-2003  19:15  2000.80.780.0   7,516,241 bytes  Sqlservr.exe
   08-Feb-2003  06:40  2000.80.765.0      45,644 bytes  Sqlvdi.dll
   26-Feb-2003  03:41  2000.80.777.0      29,244 bytes  Ssmslpcn.dll
   26-Feb-2003  03:41  2000.80.777.0      82,492 bytes  Ssnetlib.dll
   28-Feb-2003  01:34  2000.80.778.0      98,872 bytes  Xpweb70.dll

Note Because of file dependencies, the most recent hotfix or feature that contains the files may also contain additional files.

MORE INFORMATION

Error message 17883 is introduced in SQL Server 2000 Service Pack 3 (SP3). The error message can be triggered by the checkpoint condition described in the "Symptoms" section. For the error to occur and be written to the SQL Server error log, there must be at least 100 MB of dirty buffers that SQL Server flushes.

SQL Server writes this error message to the error log because of a problematic situation with SQL Server Schedulers. The error message indicates that some resource is not appropriately yielding the SQL Server Scheduler. This causes the non-yielding User Mode Schedulers (UMS) worker to run for extended periods, thereby delaying other workers on that SQL Scheduler. The server may also become unresponsive.

By default, the SQL Server 2000 checkpoint process keeps 100 pending I/O operations outstanding before it will yield to the UMS scheduler. Therefore, the I/O rate to delay the UMS scheduler has to be high, and the disk system has to respond to the GatherWrite request without COMPLETED and not I/O PENDING. It generally takes a high-end computer to dirty enough pages and to contain a large enough disk system to achieve a lengthy checkpoint.

This might have effects for less than the 60 seconds checked by the error message. Even UMS scheduler stalls at intervals of 2 or 3 seconds can cause a transactional interruption. If the computer that is running SQL Server is experiencing transaction changes during checkpoint, it is a good idea to consider the hotfix discussed in this article as a possible candidate to help reduce the effects.

Compression can also be a factor. SQL Server does not support writable, database file storage on a compressed drive. The compression algorithms disable the Write-Ahead Logging (WAL) protocol, and can affect the timing of the WriteFile calls. As such, this can also lead to stalling conditions during checkpoint.

To determine whether the system is experiencing the problem, you can monitor the performance monitor counter Checkpoint pages/sec of the SQLServer:Buffer Manager Performance object . The counter indicates the duration of checkpoint, and how many pages are involved. Also, you can add these PhysicalDisk object counters to determine if checkpoint can complete significant IO operations in a short time:

Avg. Disk Queue Length

-and-
Disk Transfers/sec