August 2008 - Posts
On my previous posts on this health monitoring series I explain to you how and why I made my own EventLogWebEventProvider and which benefits you can achieve by using this provider or by making your own.
Now I'll write about how to use this new provider in one application.
Well, almost everything have been written about this topic and Microsoft has one good article on "How To: Use Health Monitoring in ASP.NET 2.0" so I wont even try to explain all the possible scenarios and configurations, I will simply explain the standard scenario.
Typically I simply want to keep track on two Web Event types:
- the ones related to application errors
- an the others related to application life cycle
As far as I notice there are many people that simply track those from point 1.
Another decision to make is where to store these Web Events data, You do this by choosing a Web Event provider.
ASP.NET give us out-of-the-box several providers and you can make your own provider. Once your provider is done you can use it just like all the others.
In this example I will use the newly created provider.
Now that I know exactly what to track and where to store it I can update my configuration.
All health monitoring configuration data is stored in the system.web\healthMonitoring section and my standard configuration will look like:
<add name="ExtendedEventLogWebEventProvider" type="NG.Web.Management.EventLogWebEventProvider, NG.Web" source="MyEvtLogSource" />
<add name="Application Lifetime Events Default" eventName="Application Lifetime Events" provider="ExtendedEventLogWebEventProvider" profile="Default" />
<add name="All Errors Default" eventName="All Errors" provider="ExtendedEventLogWebEventProvider" profile="Default" />
Notice the source attribute from the ExtendedEventLogWebEventProvider, using it you can set which EventLog source to use.
Remember that when you create a EventLog source you can choose to create a brand new EventLog and if you choose so, all entries written by this provider in this application context will be isolated from all the others providing one easy way for visual tracking and filtering.
With the health monitoring data in place you should now be able to see the entries appearing in EventLog.
In the first post of this series I've manage to find the correct eventId for each Web Event type, and by this time the major problem has been solved, but I cannot yet write a correct entry into the EventLog.
I still have to decided the best severity type and category to apply.
If you look at entries in the EventLog generated by the default EventLogWebEventProvider you will find that they are marked mainly as Information and a few of them are also marked as Warning but no one is ever marked as Error.
Since I'm making my own provider I will take this chance to map the EventLog entry severity type according to the source Web Event.
Naturally, you can decided to map the Web Events exactly has the default EventLogWebEventProvider does.
Finally all I need is to do is set the correct category for the Web Events.
Although all Web Events belong to the same category 'Web Events', the problem is how to make the correct text appear.
The value shown in the category property is a resource string and to select the correct category value is necessary to set the exact resourcekey.
That is not so hard to do, I could have made my own resource assembly with only one resource, but I have a better way ... I will use the same resource assembly that ASP.NET use.
To figure out which assembly to use I simply used regedit.exe to look at "HK_LM\System\CurrentControlSet\Services\EventLog\Application\ASP.NET 2.0.50727.0"
What I look for are the settings used by the "ASP.NET 2.0.50727.0" EventLog source which is the source used to write the Web Events EventLog entries.
The settings I will use are the CategoryCount and CategoryMessageFile keys because they are the ones that instruct which category assembly to load.
|CategoryCount ||5 |
|CategoryMessageFile ||C:\WINDOWS\Microsoft.NET\Framework\v2.0.50727\aspnet_rc.dll |
Please note that the CategoryMessageFile may change according to operating system and ASP.NET version.
Finally, I need to know the exact resourcekey of the "Web Event" string, and since there simply 5 resources in assembly it took no time to get the following table:
Category Resource Key
|0 ||None |
|1 ||Setup |
|2 ||Uninstall |
|3 ||Web Event |
|4 ||File Monitoring |
|5 ||Admin Service |
Now I simply need to configure my EventLog source to use those same settings.
As I said in part 1, you can create an EventLog source using the EventLog.CreateEventSource method, but for setting the CategoryCount and CategoryMessageFile values you must use this specific overload EventLog.CreateEventSource(EventSourceCreationData sourceData).
That's it ... I've got everything I need and the final result look like this:
It looks similar to the usual but if you look closer you will find the differences.
What you need to keep in mind is that:
- the EventId and Category values are keep unchanged;
- the entry Type respects the Web Event type as opposing to the EventLogWebEventProvider given by ASP.NET that only uses Information and Warning;
- the EventLog source can be one of my choise (not the usual "ASP.NET x.xxx") and can change by application;
- you can choose either to use the existing EventLogs or to create a brand new one for your application(s) .
As a final note remember that with this provider you can now organize your application EventLog entries into specific EventLogs, and you may also filter them by application Source.
Download the code here.
I'm not an every day SQL Server user but I use SQL Server regularly since 7.0 version until the 2005 version (not yet tried 2008 in a serious way) and from time to time I still find some nice hidden gems.
A few days ago I needed to created a cleanup script for an application and one of the tasks was to drop all tables that match a specific name pattern.
My first thought was to use a cursor to loop or a dynamic SQL statement ...
... but this time I decided to google for some other approach, and I found the amazing undocumented sp_MSforeachtable stored procedure from the master database.
It does the same but it requires considerably less code and improves the script readability.
Below is the syntax for calling the sp_MSforeachtable SP: [more]
exec @RETURN_VALUE=sp_MSforeachtable @command1, @replacechar, @command2,
@command3, @whereand, @precommand, @postcommand
- @RETURN_VALUE - is the return value which will be set by "sp_MSforeachtable"
- @command1 - is the first command to be executed by "sp_MSforeachtable" and is defined as a nvarchar(2000)
- @replacechar - is a character in the command string that will be replaced with the table name being processed (default replacechar is a "?")
- @command2 and @command3 are two additional commands that can be run for each table, where @command2 runs after @command1, and @command3 will be run after @command2
- @whereand - this parameter can be used to add additional constraints to help identify the rows in the sysobjects table that will be selected, this parameter is also a nvarchar(2000)
- @precommand - is a nvarchar(2000) parameter that specifies a command to be run prior to processing any table
- @postcommand - is also a nvarchar(2000) field used to identify a command to be run after all commands have been processed against all tables
As you can see, this stored procedure offer us some flexibility, but for the most common uses you will only use one or two of them.
Back to my problem, drop all tables with a specific naming pattern, I ended up using a script just like this:
declare @appName varchar(128)
declare @mycommand varchar(128)
declare @mywhereand varchar(128)
set @appName = 'xpto'
set @mycommand = 'drop table ?'
set @mywhereand = 'and o.name like ''' + @appName + '__Log__%'' escape ''_''
print 'Dropping all tables belonging to ' + @appName + ' application ...'
@command1 = @mycommand,
@whereand = @mywhereand
What I'm saying here is that the command 'drop table' should be executed for every table that match the criteria name like 'xpto_Log_%'.
As you can see its fairly simple and clean and this is just the top of the iceberg.
For more detail about sp_MSforeachtable go here and here.
The ASP.NET health monitoring enables you to add instrumentation to Web applications by using the so called Web Events. These Web events give us information about health status.
You can configure health monitoring by setting events and providers in the healthMonitoring section.
Naturally, ASP.NET provide us with a few out-of-the-box providers such as the EventLogWebEventProvider.
As many of you may have already notice, when using the EventLogWebEventProvider the events are added to the Application EventLog with the following source pattern:
ASP.NET "framework version"
If you are using ASP.NET 2.0 the source will look similar to "ASP.NET 2.0.50727.0 ".
You can imagine what happen when a server hosts several web applications ... you can't easily figure which application raised a web event because you can't apply a filter to do that. To figure it out you must inspect the eventlog entry data.
If you think this is wrong and that Microsoft should do something about it, go here and here, vote and comment.
What can you do to overlap this? Well you can create your own EventLogWebEventProvider that allows you to specified which Source to use.
Doing such a provider is fairly simple but lead us to THE problem: which eventId to use when creating the EventLog entry?
What? Why is this a problem? are you saying.
Well, lets start all from the beginning ... you want to create your own provider so you can specify the EventLog source but you certainly desire to keep the remaining settings unchanged so that monitoring applications that track the EventLog entry for well known eventIds still working fine.
The problem is that Microsoft don't expose the algorithm used to created the eventId from the WebEvent data, and this way we can only guess which eventId to use.
If you look at EventLogWebEventProvider.ProcessEvent method you will find the following code:
int num = UnsafeNativeMethods.RaiseEventlogEvent((int) type, (string) dataFields.ToArray(typeof(string)), dataFields.Count);
This is your black box, no source or information is available.
To guess which eventId is used for a specific Web Event I created a small page that raises all known Web Events.
I found that even with all known Web Events configured to use EventLogWebEventProvider almost half of them don't appear in EventLog, but those that have an EventLog entry made me speculate that eventIds are sequential and follow the classes hierarchy. Here are the results:
Please note that I'm considering that no two different Web Events share the same eventId.
If you believe the assumptions made are correct you can now start coding your provider.
Remember that you must create the EventLog source before use it. You can do this by using the EventLog.CreateEventSource method.
Recently, while digging on ASP.NET 2.0 Health Monitoring I found a bug in the EventMappingSettingsCollection.Contains class method.
I was trying to check if an event mapping already exists but every time I try it the following exception was thrown:
System.NullReferenceException was unhandled by user code
Message="Object reference not set to an instance of an object."
at System.Web.Configuration.EventMappingSettingsCollection.GetElementKey(ConfigurationElement element)
at System.Configuration.ConfigurationElementCollection.GetElementKeyInternal(ConfigurationElement element)
at System.Configuration.ConfigurationElementCollection.BaseIndexOf(ConfigurationElement element)
at System.Web.Configuration.EventMappingSettingsCollection.IndexOf(String name)
at System.Web.Configuration.EventMappingSettingsCollection.Contains(String name)
I opened a bug in connect, so, if you think this is important go there and vote.
And if you think this one is not a common error ... well ... this is the second bug I found regarding Contains method from a collection class.