I've got a WCF client communicating with a WCF service using [DataContract] types, and I'm getting a serialization error:
The formatter threw an exception while trying to deserialize the message: There was an error
while trying to deserialize parameter http://www.example.com/2007/09/Example:ExampleResult.
The InnerException message was 'Deserialized object with reference id 'i3' not found in
stream.'. Please see InnerException for more details.
Normally, I'd simply crank up the tracing and see exactly what happened, but in this case, I can't get the offending (response) message to appear in the log.
My configuration looks like this:
<system.serviceModel>
... (more stuff)
<diagnostics>
<messageLogging logEntireMessage="true" logMalformedMessages="true"
logMessagesAtServiceLevel="false" logMessagesAtTransportLevel="true"
maxMessagesToLog="10000" maxSizeOfMessageToLog="81920000" />
</diagnostics>
</system.serviceModel>
<system.diagnostics>
<trace autoflush="true" />
<sources>
<source name="System.ServiceModel" switchValue="Information, ActivityTracing"
propagateActivity="true">
<listeners>
<add name="wcf_listener" />
</listeners>
</source>
<source name="System.ServiceModel.MessageLogging">
<listeners>
<add name="wcf_listener" />
</listeners>
</source>
</sources>
<sharedListeners>
<add name="wcf_listener" initializeData="tracelog.svclog"
type="System.Diagnostics.XmlWriterTraceListener" />
</sharedListeners>
</system.diagnostics>
In the resulting log file, I get the outbound message logged, and then the exception logged. I never see the incoming message. What am I doing wrong here?
It seems that the problem related to your flooded data between the service and the client.
Please refer to my post here
Related
I'm logging messages for WCF web service using System.Diagnostics System.ServiceModel.MessageLogging with configuration below. And I extended
TraceListener class as:
public class FormattedTraceListener : TraceListener
{
static readonly Logger logger = new Logger();
public FormattedTraceListener()
: base(string.Empty)
{
}
public override void TraceData(TraceEventCache eventCache,String source,TraceEventType eventType,**Int32 id,**Object data)
{
//....
<system.diagnostics>
<sources>
<source name="System.ServiceModel.MessageLogging" >
<listeners>
<add name="pretty" />
<remove name="Default"/>
</listeners>
</source>
</sources>
<sharedListeners>
<add name="pretty" lockItem="true" type="LoggingTest.FormattedXmlWriterTraceListener,LoggingTest" />
</sharedListeners>
<trace autoflush="true" indentsize="4">
</trace>
</system.diagnostics>
....
<diagnostics >
<messageLogging
logEntireMessage="true"
logMalformedMessages="false"
logMessagesAtServiceLevel="true"
logMessagesAtTransportLevel="false"
maxMessagesToLog="-1"
maxSizeOfMessageToLog="134217728">
<filters >
<add xmlns:s="http://schemas.xmlsoap.org/soap/envelope/">/s:Envelope</add>
<add xmlns:s="http://www.w3.org/2003/05/soap-envelope">/s:Envelope</add>
</filters>
</messageLogging>
<endToEndTracing propagateActivity="true" activityTracing="true" messageFlowTracing="true" />
</diagnostics>
</system.serviceModel>
TraceData method fired twice (for request and response) for each service call.
However, when method is hit, the id parameter is always 0.
I need a unique id in logging text to match request and response log. How can I specify that?
Your approach is not right. The TraceData(int id) field is not a unique id for the WCF request. It's the application defined code for the event type.
Just use the standard WCF trace classes and view correlated traces. If I understand what you are trying to accomplish in matching up request to response, the default behavior already meets your requirements.
In a .Net 4.0 web-service I am using trace autoflush to write to a log file.
By adding the following in the web.config:
<trace autoflush="true" >
<listeners>
<add name="TextWriter"
type="System.Diagnostics.TextWriterTraceListener"
initializeData="trace.log" >
</add>
</listeners>
</trace>
Am trying to find a way to log only Trace.TraceError("error info") and exclude Trace.TraceInformation("some verbose debugging stuff") without altering my code and just changing the web.config?
The information I have found on MSDN shows how this can be done by adding code that calls Trace.Flush() and adding sources, switches and sharedlisteners, however I would like to continue using auto-flush and not alter the code.
Many thanks in advance :)
Old answer:
This appears to be impossible. Trace auto-flush does not have the
capacity to have it's level set in the web.config.
Update:
The listener may have a filter applied as follows
<configuration>
<system.diagnostics>
<trace autoflush="true" >
<listeners>
<add name="TextWriter"
type="System.Diagnostics.TextWriterTraceListener"
initializeData="trace4.log" >
<filter type="System.Diagnostics.EventTypeFilter" initializeData="Warning" />
</add>
</listeners>
</trace>
</system.diagnostics>
Note that the initializeData is a string and may take the values of Warning or Error. This filters the trace output accordingly.
Many thanks to Josip for following up on this question.
Maybe you can use "switchValue" for this purpose like this:
<system.diagnostics>
<trace autoflush="true">
</trace>
<sources>
<source name="SCInterface" switchType="System.Diagnostics.SourceSwitch" **switchValue**="All">
<listeners>
<remove name="default"/>
<add name="HSInterface"
type="XYZ.Diagnostics.CyclicTextTraceListener, XYZ.Base3x"
initializeData="D:\Trace\HSR.HSInterface.Trace.log, Size, 10, 5"
traceOutputOptions="DateTime,ThreadId,Callstack" />
</listeners>
</source>
</sources>
</system.diagnostics>
For switch value you would put Warning or Error...
I have a web application that I'm developing that makes a lot of HttpWebRequests. In order to make debugging them easier, I've set up the following in my web.config;
<system.net>
<defaultProxy>
<proxy proxyaddress="http://127.0.0.1:9999" />
</defaultProxy>
</system.net>
This allows all of my HttpWebRequests to proxy through Fiddler. The problem is I need to have Fiddler running in order to have my app work correctly.
Ideally, I would like to have it proxy through Fiddler when Fiddler is running, and not proxy at all when Fiddler is not running without having to change my web.config each time.
There are a few options.
First, you can set the relevant Proxy property of the relevant objects inside your code directly instead of falling back to the configuration XML; you can then selectively control the use of the proxy based on any factor you like.
Alternatively, you could try setting the scriptLocation attribute to point at http://localhost:8888/proxy.pac and use Fiddler's about:config to set fiddler.proxy.pacfile.usefileprotocol to false and tick the Tools > Fiddler Options > Connections > Use PAC Script box.
Have you considered an another approach, by enabling trace listeners on System.Net? It is not as comfortable as using Fiddler, but it might be enough for sporadic debugging. For message logging, just System.Net should be sufficient, but there are more.
<system.diagnostics>
<trace autoflush="true" />
<sources>
<source name="System.Net">
<listeners>
<add name="System.Net"/>
</listeners>
</source>
<!--<source name="System.Net.Sockets">
<listeners>
<add name="System.Net"/>
</listeners>
</source>-->
<!--<source name="System.Net.Cache">
<listeners>
<add name="System.Net"/>
</listeners>
</source>-->
</sources>
<sharedListeners>
<add
name="System.Net"
type="System.Diagnostics.TextWriterTraceListener"
initializeData="System.Net.trace.log"
/>
</sharedListeners>
<switches>
<add name="System.Net" value="Verbose" />
<add name="System.Net.Sockets" value="Verbose" />
<add name="System.Net.Cache" value="Verbose" />
</switches>
</system.diagnostics>
I have a certain framework of code, and I have a TraceListener defined for two reasons:
Back-compatibility with a lot of the old logging that was done via Trace.Write until we update it, and
It's nice to be able to instrument the other assemblies our code references if we need to.
However, I have one assembly (not ours) that logs a lot of pointless data that doesn't help us debug anything. How can I turn off tracing for this one assembly (or, alternately, the facade project we built around it), while leaving it on for the rest of the application?
I've tried various flavors of configuration in our facade project, usually looking like the following, to no avail. I've tried adding <remove> elements that match the <add> elements which setup the logging in the first place, tried <clear>ing them, setting <trace enabled="false"> and at least three other attempts. Thanks for any help you can provide!
<system.diagnostics>
<trace autoflush="true" indentsize="4">
<listeners>
<clear/>
</listeners>
</trace>
<switches>
</switches>
</system.diagnostics>
You can write your own trace filter, for use with your TraceListener. Inside this filter you can look for your assembly in stackTrace and turn off event tracing.
In my case I wrote filter (see: DotNetOpenAuthFilter) based on EventTypeFilter, which filters events only from the DotNetOpenAuth library.
Then connect the filter to the listener in the web.config:
<configuration>
<system.diagnostics>
<trace>
<listeners>
<add name="console" type="System.Diagnostics.ConsoleTraceListener" >
<filter type="Common.Log.DotNetOpenAuthFilter, Common" initializeData="Warning" />
</add>
</listeners>
</trace>
</system.diagnostics>
</configuration>
Use TraceSource.
Initialize it in your trace source.
TraceSource logger = new TraceSource("Class1");
Call it from critical points in code:
logger.TraceInformation("Hello from Class1");
Be sure to edit your application configuration:
<system.diagnostics>
<trace autoflush="true"/>
<sources>
<source name="Class1" switchName="Class1Switch" switchType="System.Diagnostics.SourceSwitch">
<listeners>
<add name="console"></add>
<add name="csv" />
<!-- or you can add your own listener here -->
</listeners>
</source>
</sources>
<switches>
<add name="Class1Switch" value="Information" />
</switches>
<sharedListeners>
<add name="console" type="System.Diagnostics.ConsoleTraceListener" />
<add name="csv" type="System.Diagnostics.DelimitedListTraceListener"
delimiter="|" initializeData="d:\data\tracing\trace.log"
traceOutputOptions="Timestamp, ThreadId, LogicalOperationStack, DateTime, ProcessId">
</add>
</sharedListeners>
</system.diagnostics>
If say, you want to only log errors, change the switch:
<add name="Class1Switch" value="Error" />
To switch it completely off:
<add name="Class1Switch" value="Off" />
I have a working WCF service that I wanted to hook into so I can look at the logs. So, I added a .config file, and now my service will not start at all.
In my code I do the following:
private ServiceHost _myHost;
....
_myHost= new ServiceHost(typeof (MyType), new Uri(baseUri));
_myHost.AddServiceEndpoint(typeof (myInterface), new NetNamedPipeBinding(), hostName);
_myHost.Open();
And, as mentioned, this works when I do not have a .config file. However, to view the logs I have to add a .config file that looks like this:
<system.diagnostics>
<sources>
<source propagateActivity="true" name="System.ServiceModel" switchValue="Verbose,ActivityTracing">
<listeners>
<add type="System.Diagnostics.DefaultTraceListener" name="Default">
<filter type="" />
</add>
<add initializeData="c:\logs\Traces.svclog" type="System.Diagnostics.XmlWriterTraceListener"
name="traceListener" traceOutputOptions="LogicalOperationStack, DateTime, Timestamp, ProcessId, ThreadId, Callstack">
<filter type="" />
</add>
</listeners>
</source>
</sources>
</system.diagnostics>
<system.serviceModel>
<diagnostics>
<messageLogging logMalformedMessages="true" logMessagesAtServiceLevel="true"
logMessagesAtTransportLevel="true" />
<endToEndTracing activityTracing="true" messageFlowTracing="true" />
</diagnostics>
</system.serviceModel>
The error I now get in the logger is Failed to open System.ServiceModel.ServiceHost and it occurs at the instance creation of my ServiceHost. So, why does adding a logging config make any difference?
UPDATE
The logger is the ServiceTraceViewer, that is where I am seeing the error once I throw in the config. So, the logging is working, it is just breaking the functionality of the code. I read somewhere that if I use a .config that I MUST have an endpoint in the config. Is that true, and if so, is there a workaround to this since I am doing that via code