Page 2 of 2 FirstFirst 12
Results 11 to 12 of 12
  1. #11
    Senior Member
    Join Date
    Oct 2001
    You have never had problems with your Win2K cluster corrupting Exchange? Microsoft's Exhcnage project manager himself has even apologized for it in public. In fact they no longer recomend an Exchange 2000 Cluster on windows 2000 unless it's very specific in configuration that they trust and that is not Active/Active or even Active/Passive it just does not work and this "sucks ass" in that situation. I guess a file share or load balancing wouldn't be much of a problem or a non-cluster aware virtual server with data stored on the cluster... and I am not familiar with clustering SQL, that might be king.
    I highly doubt that anybody at MS has said that Win2k and Exchange2k no longer supports clustering. I work with MS on a daily basis, I actually know several of the critical problem resolution technicians on a personnal basis because of the close interaction we have had with them over the last 3 years. I've not had any problems with MS clustering, specifically disk issues since the clusdisk.sys patches were released almost 3 years ago. Even then we were not losing data, we were just losing disk signatures, which is easily corrected. Active/active clusters have never been supported in Exchange2000 as they have a lot of memory issues when the concurrent connection number goes over 1000.

    One of the big problems with clustering is the lack of great documentation for it. I have about 40 pages of technical procedures that I have written that all revolve around using a win2k cluster and exchange. Most of the information came directly from MS, but I've not seen most of the stuff that I have properly documented on the MS technet site. The order in which you start/stop/failover nodes can impact your disk structures. There is a pretty good chance that if you bounce both nodes of a cluster at the same time they will corrupt the disk signatures in the cluster metabase. So yeah, there is a lot of potential to mess things up, but not if you know the product well.

    MS does not official support any clustering configuration unless it is very specific in nature, that is nothing new. There are certain firmware revisions, and hardware devices that will not work. If you are not running pretty much the latest and greatest firmware and BIOS revisions on all of your networking components and SAN components, which all have to be officially supported, you will have disk issues. But that is not to say that the product cannot be made stable, or is not recommended. I've never used Dell servers in clusters. Compaq/HP is the preferred vendor for datacenter and clustering and we use their hardware exclusively. Is the product extremely scalable and stable out of the box? Not really, but if you are going to try and run more than 5000 clients per cluster with over 1TB of data and the need to back all of the up in a reasonable amount of time you had better know the product inside and out. Running a default configuration in that instance would not make sense at all. It's not that difficult to tweak the disk and memory configurations in such a manner to make the system stable.

    I know the exchange 2000/2003 product manager from being a member of the Exchange 2000 Joint Development partner program. We are still a member of JDP although we did not participate in the early adopter program for 2003 like we did for 2000. If you are a member of the Exchange JPD/dogfood program let me know and I will tell you who I am as I post on that list sometimes.

  2. #12
    Senior Member RoadClosed's Avatar
    Join Date
    Jun 2003
    MS does not official support any clustering configuration unless it is very specific in nature, that is nothing new.
    That's all I am saying. I was SOLD on this thing by Microsoft and Dell for it's ability to keep juggin' along with failover. Both Claiming it was tested and "certified" The issue is, it's so touchy it almost never fails over and stays alive. And that is the sole reason I got it. Sure I can be very careful and fail the nodes over properly but what good is that? I grindge at any sort of minor error on the thing and it's Supposed to be fault tollerent. LOL. That is why you buy a cluster outside of distributing work load. I am with you on the lack of documentation. There is some more out now but back 2 years ago I had to feel my way in the dark. So I have to wonder if MS clustering 1. Sucks and 2. If other clustering models are better?

    I am not on the list, but I did speak face to face with the product manager over Exchange 2003 and Outlook 2003 at a conference in Denver. It was his own words i was reiterating.

    Oh and I do spend a lot of time taking it down (note not fault tolerance) and updating the firmware of a gazillion devices to keep up with it. If you don't take it down and update 1 node then chances are very good something will get corrupted so you take the whole thing down update 1 node, power it off, update the other then power the controlling node back up, then the passive and CROSS your fingers. That in my humble opinion is a waste and denys the very reason why clustering is useful outside of load balancing.

    We are on the same page just a different book.
    West of House
    You are standing in an open field west of a white house, with a boarded front door.
    There is a small mailbox here.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts

We have made updates to our Privacy Policy to reflect the implementation of the General Data Protection Regulation.