Welcome to EMC Consulting Blogs

Sometimes the burndown chart for a sprint may not look the way you expect, even after sorting out all the hours and the remaining effort involved.  If they do not look correct then there is problem something wrong with the data warehouse of the TfsWarehouse database.

The first thing to do is to see how recently the warehouse was updated.  The way to do this is to log into the SQL server of TFS (erither using RDP or direct from the SQL Management Suite) and run the following query on the TfsWarehouse database:

use [TFSWarehouse]

select top 1 __LastUpdatedTime from dbo.[Work Item History] with (nolock) order by __LastUpdatedTime desc

select * from _WarehouseConfig

The output will include something similar to the following:

__LastUpdatedTime

-----------------------

2007-08-06 17:58:37.560

This was a scenario that happened at one of our clients and we were investigating this problem on Wed 15th August.  This showed that the warehouse had not updated for at least 6 days!

The immediate reaction might be to rebuild the Warehouse, indeed this is what I did, however this does not address the issue.  The way to fix this problem is to check the analysis service of SQL running on the data-tier of the TFS server.  It may have stopped running, in which case the reason why it stopped needs to be investigated, however if it is running restart it.

Once the service has restarted check the System Event log, it is possible that although it has started it is moaning about some files in the filesystem.  This is the error that we got:

Event Type:        Error

Event Source:    MSSQLServerOLAPService

Event Category:                (289)

Event ID:              18

Date:                     15/08/2007

Time:                     14:20:45

User:                     N/A

Computer:          TFS-SERVER

Description:

Errors in the metadata manager. An error occurred when instantiating a metadata object from the file, '\\?\C:\Program Files\Microsoft SQL Server\MSSQL.2\OLAP\Data\TfsWarehouse.0.db\Team System.12604.cub.xml'.

Using explorer on the server go to the location specified in the log file.  In our case there were a bunch of 0 byte XML files that had created; strangely these files were compressed.  Any XML file that is corrupt or does not contain anything will cause problems for the update process.  Delete these files (as specified in the event log) and restart the service which should now start cleanly.

The last step in the process is to get the TFS system to update the warehouse.  Use the web service for TFS to invoke this, BUT bear in mind that it could well take a long time and on a busy TFS server it maybe better to do outside of hours - although remember that TFS itself will try to updated the warehouse every hour by default.

One of the things that I loved about my career is that I have been able to stay multi disciplined with Microsoft, UNIX and Open Source products; and I very pleased to say that i have been able to continue at Conchango.

 All technologies have a place in the market, and whilst I sometimes play devils advocate in the old Windows vs. UNIX debate I firmly believe that there is a no clear cut answer.  What is good for one solution may not be good for something else.  It is this understanding of the different systems that has given me a much broader view of technologies and has helped many a time in projects that I have worked.

So how has this all helped?

Personally I love to play with Linux and find out what extra things it can do, and over time i have learnt that it can do some truely spectacular things that other systems can do but bormally cost the earth.

Recently I was on a project that used HP servers and Windows 2003 Server for a development and test environment.  There are sme monitoring management tools that come with HP servers, but as the name suggests it is aimed at HP servers and there were a couple of servers that were not HP based so it would be (near) impossible to manage.

This is where Linux comes in.  My favorite distribution is Gentoo (www.gentoo.org) as it is a build from scratch 'distro' and you can have as much or as little as you like.  Using this as a platform and installing some truely brilliant open source applications:

• Apache 2
• PHP 5
• MySQL 5

It then became a very simple matter of monitoring any system you like, be it a network switch, router, Windows or UNIX server.  The key thing here is that they can all talk the same language, Simple Network Management Protocol (SNMP).

Monitoring

OK so I have sold you on the platfor, but how does it all work?  The answer lies in using a PHP module and some nifty Perl scripts - all of which are distributed uinder the GPL licence.

1. Cacti - www.cacti.net
2. NetSNMP

Cacti uses the cron system to periodically poll servers for status, anything from Network usages to disk space.  This is then pulled back and charted using the superb RRDTool (www.rrdtool.org).  It gets the information using the NetSNMP package.

There are many more systems that can be added to this, such as Nagios (www.nagios.com), SmokePing (http://oss.oetiker.ch/smokeping/), Nessus (www.nessus.org) to name but a few.

Learning Curve

I think one of the biggest problems people have with Linux and open source software is that they do not know how to use it and thus are fearful which makes them shy away from these systems.  In reality once the base system has been setup most of the applications have extremely friendly and easy to use web interfaces.

The draw for me is that I can make Linux do (almost) anything I want it, albeit using the experience of other people - but this is the great thing about Open Source software.

I know this is an extremely small insight into why I like Linux and what it can do and I will do and I wille xpand on it in the next few weeks, but my primary reason for this blog is that I balled over by the fact that I am able to continue using this technlogy at Conchango.  It is great that I share information about the diversity of Linux and showcase projects and applications that I use.  We have a great community dedicated to Open Source and J2EE applications which is full of people that have similar views.

Ok so this is the geek spot, but please look at Michelle Flynn's blog (our Recruitment Manager) for a taste of the personal side of Conchango