THE SQL Server Blog Spot on the Web

Welcome to SQLblog.com - The SQL Server blog spot on the web Sign in | |
in Search

Ben Miller

Cluster bug rears its head once more

Back in July 07, I found and verified a bug in SQL 2005 and Windows 2003 when clustering SQL Server. From previous cluster setups, I found that if I was given a Windows 2003 R2 x64 Server with SP2+ on it, then the cluster went well no matter if the machine name was lower case or not, but the key to this entire bug is that if you go into Cluster Manager and add the nodes to the Cluster, and you see a lowercase name in the list of cluster nodes, SQL will ALWAYS fail.  I have verified it so many times, that I am convinced.  But this only happens when you are on 2003 with SP1.  You can even verify that something is wrong by finding the registry keys that have the server name in them and see that some are lower and some are upper.

There is much more to this story and one day I will dialog it all out further than I have, but with Windows 2008 on the horizon and SQL 2008, hopefully these kinds of cases will not find their way into the mainstream.  It must be only our organization because the connect item above has not had much traction.

Published Wednesday, April 09, 2008 10:49 AM by dbaduck
Filed under:

Comment Notification

If you would like to receive an email when updates are made to this post, please register here

Subscribe to this post's comments using RSS

Comments

 

AaronBertrand said:

I spent about a year with PSS, off and on, trying to repro / resolve an issue where a SQL Server 2005 cluster had SP1 "sort of" installed.  It took on one node, but somehow rolled back on the second node.  Suddenly you could tell which node you were on simply by selecting @@VERSION (the "bad node" still said 1399).  Due to another bug I spent a lot of time with PSS on, relating to tempdb memory leaks, but that is thankfully fixed now (3161 I think), we were manually failing over every weekend.  Except we had to fail over and then fail back to ensure that the "good node" was active.  When SP2 came out, I thought, finally, we can stop this nonsense and not be so concerned about which node is active.  This problem was still not resolved.  As a part of an unrelated hardware migration, I finally retired that cluster and moved the application to a new cluster... on which both nodes are currently running 3228.  What a nightmare leading up to that though... you never knew when a failover was going to cause bugs to resurface on the node that was still at RTM...

April 9, 2008 9:26 PM
 

alphatross said:

MS still has a long way to go with SQL on MSCS Clusters. The company I work for has a Win 2003 R2 \ SQL 2005 Enterprise x64 Active-Active 2-Node Cluster - once you get it installed it's great, but a *lot* of pain to get there, and once there you have to be *so* careful doing any changes. Just document absolutely all the install steps to make it reproducable, and automate installation tasks as much as possible to avoid weird issues. E.g. if one Node has a Network Connection named "Heartbeat 1", and another Node's Network Connection was mistakenly labled a space after the name "Heartbeat 1 ", you can get all the way through setting up MSCS, installing SQL 2005, but SP2 will continually fail.  Same issue if you are logged into both Nodes at time of SP2 install can happen. I hope SQL 2008 \ Windows 2008 is better!

April 10, 2008 7:11 AM
 

Manish K said:

Hi Ben,

I am very thankful for this valuable article, how ever I liked the way u did to get the dependencies of the table and then script as per the order, and I would like to do the same with Views of my database, where some of the views are dependent on other views, I tried using

DependencyTree tree = scr.DiscoverDependencies(view, true);

DependencyWalker depwalker = new Microsoft.SqlServer.Management.Smo.DependencyWalker();

               DependencyCollection depcoll = depwalker.WalkDependencies(tree);

but it didnot worked for me, Please advice.

Thanks

Manish

manishwc@gmail.com

October 26, 2008 3:37 AM

Leave a Comment

(required) 
(required) 
Submit

About dbaduck

Ben Miller is a Senior Database Administrator for HealthEquity in Draper, UT. He has been working with SQL since SQL Server 6.0 (1998) and has had a variety of roles in his career, including SQL Support and MVP Lead at Microsoft.
Powered by Community Server (Commercial Edition), by Telligent Systems
  Privacy Statement