Microsoft KB Archive/255198

= Cannot Change DCOM Call Timeouts in Windows NT 4.0 =

Article ID: 255198

Article Last Modified on 11/1/2006

-

APPLIES TO


 * Microsoft Windows NT Server 4.0 Standard Edition

-



This article was previously published under Q255198



SYMPTOMS
Distributed Component Object Model (DCOM) calls may take a long time to time out if the network is down, or if the server is unavailable.



CAUSE
This behavior occurs because DCOM does not give access to the underlying remote procedure call (RPC) binding handle, which makes it difficult to configure the RPC timeout.



RESOLUTION
To work around this behavior, use the following method:

When you develop code that uses DCOM on Microsoft Windows NT 4.0, Microsoft Windows 95, or Microsoft Windows 98, you can make use of multithreading to give up on a DCOM call before the default timeout. You can spawn another thread that makes the DCOM call, and have the original thread wait for the desired amount of time. The original thread can 'give up' on the spawned thread, and proceed as if the call had timed out. There are a few things to consider when implementing this solution:


 * If the original thread wants to really 'forget' about the thread that is making the DCOM call, it must close the handle to the thread, so that when the original thread ends, the thread object is deleted.


 * The original thread might give up and proceed as if the call did not succeed, but then network conditions might improve so that the call does succeed. If the call modifies the state of the server, this can be a problem. It might be possible for the background thread to 'clean up' if it has been abandoned. This type of solution is best for calls that do not modify the state of the server.


 * Even though the original thread gives up on the thread that made the DCOM call, the thread is still running until the DCOM call times out. This means that the process cannot end normally because it will wait for the thread to end. You can use TerminateThread in this situation, but it is not recommended.



MORE INFORMATION
The actual DCOM timeout depends on a number of factors, including which network protocol DCOM is using and whether the server or network went down before or during the call. DCOM makes use of RPC, and although RPC does allow some amount of control over call timeout, the programmer cannot normally take advantage of it because DCOM does not give access to the underlying RPC binding handle.

DCOM on Microsoft Windows 2000 addresses this problem by providing features such as asynchronous calls and call cancelling. If you want to cancel DCOM calls in Windows NT 4.0, Windows 95 or Windows 98, you can try using 'apartment model' threading and implementing IMessageFilter; however, this does not work well for all cases and is not recommended.

These timeouts apply to DCOM in various situations when it is using the User Datagram Protocol (UDP):


 * If the network or server is down before the start of a call, the call will time out in 32 seconds.


 * If the network or server goes down during a call, the timeout depends on how long the call lasted. The maximum timeout is between 64 and 96 seconds.


 * Timeouts are longer when DCOM is using Transmission Control Protocol/Internet Protocol (TCP/IP).

Keywords: kbprb KB255198

-

[mailto:TECHNET@MICROSOFT.COM Send feedback to Microsoft]

© Microsoft Corporation. All rights reserved.