INF: How to Troubleshoot SQL Server Communication Error 17832 (169521)

SUMMARY

Under certain situations, Microsoft SQL Server may log the following error either in its errorlog or in the Windows NT event log:

17832 Unable to read login packet(s).

This error occurs if a client starts to connect but never successfully completes the attempt because of a client operating system, application, or network failure. Wide area networks (WANs) are more susceptible than local area networks (LANs) to this type of problem, because such networks often have many routers and other network devices that can delay data packet delivery and/or cause other difficulties in the delivery of data packets due to improper packet fragmentation or misrouting.

However, the very occurrence of these errors in the errorlog or event log does not necessarily indicate any real problems with your SQL Server connectivity. Even in normal operations, this error may still occur at times, due to the unreliable nature of networks, especially wide area networks. In addition, improper use of applications (such as restarting the computer in the middle of a login process) can also cause this error to occur.

To determine whether or not you are facing a real problem, you may want to check the following:

Check with the end users to see whether they have experienced any difficulties in making connections to SQL Server. If nobody reports any connectivity problems, you can usually ignore these errors.
Check the frequency of this error in the errorlog. If it occurs very frequently, it indicates some potential problems, even if no one is complaining about any connectivity problems.

If you have determined that you have a real problem, you may want to apply the additional troubleshooting methods described in the remainder of this article.

MORE INFORMATION

Errors of type 17832 can generally be classified into two categories:

Frequent, repeated occurrence of these errors. If you examine the timestamps of these errors, you may find that they usually occur in regular intervals, typically one or more occurrences per second. The errorlog typically is full of these errors. This is mostly caused by client-side problems such as operating system or network software. The repeated occurrence of these errors is caused by the connection retries built into the client-side DB-Library or ODBC, which does not stop until the client application times out.
Intermittent, random occurrence of these errors. In the errorlog, you will find these errors occurring once in a while without any regularity. If nobody has reported any connectivity problems, you can generally ignore these errors. If someone did report connectivity problems to SQL Server, you need to use the timestamps to determine whether these reported problems correspond to the errors of type 17832 in the errorlog.

Because SQL Server does not know the client's address or host name yet when the error occurs, this error message cannot provide any information about which client causes the error. Therefore, it can be very time consuming to troubleshoot this error, especially if it occurs only intermittently. If you cannot identify which client computer is likely causing the error, you may have to use network sniffing tools such as Network General's Sniffer or Microsoft's Network Monitor to capture the sniffer trace for further analysis. The best approach is to check the following known issues first, and contact Microsoft SQL Server Support if none of following causes matches your case.

Error 17832 can be caused by:

A Windows 95 Winsock issue. A 16-bit Windows application using TCP/IP sockets may encounter this problem under Windows 95 in a slow network environment. The error is repeated numerous times in the errorlog. For more information, see the following article in the Microsoft Knowledge Base:
164516 : INF: Error 17832 Caused by Blocking Timeout in 16-Bit Sockets
Some anti-virus software on named pipes clients. Typically, when you configure anti-virus software to scan all files, named pipes (which are considered files) may be affected adversely. In this case, named pipes may be opened in invalid mode, causing these errors on the server side. To correct this problem, uninstall the anti-virus software to verify that this is indeed the case, and contact the software vendor for information on how to disable the auto-scanning of named pipes.
The use of the 32-bit ODBC driver for 16-bit Windows application under Windows 95. The 32-bit ODBC driver for SQL Server is not supported for use with 16-bit applications under Windows 95. The error occurs repeatedly in the errorlog for the duration of client login. For more information, see the following article in the Microsoft Knowledge Base:
140697 : INF: Win16 ODBC Applications in a Win32 Environment
Insufficient system resources on the client side. If a client's request for opening connections fails due to insufficient resources on the client computer, the open connection function returns with an error status, and no login packets are sent as a result, causing these errors. For more information on this known issue, see the following article in the Microsoft Knowledge Base:
161169 : BUG: "Too Many Open File Handles" Error Using Named Pipes
Insufficient system resources on the server side. If SQL Server, the Windows NT operating system, or network services cannot allocate enough resources to complete new connections, the connections are reset by server, thereby causing these errors to occur. For more information on this known behavior, see the following article in the Microsoft Knowledge Base:
154628 : INF: SQL Logs 17832 With Multiple TCP\IP Connection Requests
A client interruption in the middle of a login process. If a client application does not wait until the login process finishes, and either the application is stopped or the client computer is restarted, this error may occur each time such an abnormal termination happens.
An ODBC application that times out. There is a small window of opportunity in the SQL Server ODBC driver that may cause these errors when the ODBC application times out. Typically this error occurs when the application's timeout setting is just a small bit longer than the actual time it takes to complete the login process.
Some routers' configurations in a routed network not allowing bigger data packets to go through properly. This is often caused by incorrect router configurations, "black hole" routers, or unreliable networks such as RAS connections. For TCP/IP networks, you can generally use the TCP/IP PING utility with the "-l" parameter to test this case, but more complex issues may require the use of network sniffing tools to analyze the bottleneck. For more information on some relevant issues, see the following articles in the Microsoft Knowledge Base:
159211 : Diagnoses and Treatment of Black Hole Routers

136970 : PMTU Black Hole Detection Algorithm Change for Windows NT 3.51
The Network being too slow or unreliable. In a WAN or heavily routed network, it may take many retransmissions to deliver a large data packet successfully. If such a delivery of SQL Server login packets takes more than 45 seconds, this error will occur every time. The only solution in this case is to tune or upgrade the network; otherwise, the network is too slow for any meaningful SQL Server applications.