Updating BIN to live server 1

[jojo11]

Hi:

We maintain an enterprise web site that runs on Windows 2003 using a server farm of 5.

It is IIS6 with .NET 1.1

When we do a replication of changes into production and those changes involve an update to a dll, there are times when they have to restart IIS to get the .NET portions of the site working again. Out of the 5 servers, it is typical that 3 of them will fail for any aspx page that is called that utilizes the dll's.

Our Systems group is coming down hard on us (the development group) saying this is bad and we have to fix it.

However I thought I remember reading in some whitepaper that Microsoft recommends taking servers out of a farm if a bin change is requested as there may be issues when a request is being processed by a dll at the same time a new one is trying to overwrite the old one.

We have told them they should do this to avoid the problem but they continue to replace bin files on the server while it is in the farm and live. They say that it takes too long to take one out of the farm, copy, put back in the farm and do it five times.

I am looking for ammunition to take to them to prove that they should be taking the server out of the farm when we make these updates.

Does anyone have anything to share on this that may help our cause?

Also if it is known that there is an issue replacing live DLL's then are there any updates to .NET 2.0 that help resolve this?

Thanks!!

-------------------------------------------
Ummm, we have a bit of a problem here....

 

[BoulderBum]

You can see here that Microsoft talks about taking the app offline:

http://msdn2.microsoft.com/en-us/library/f735abw9.aspx

Preparation

Use these procedures if you are deploying a large Web site and you want to take the target Web site offline while you are copying files.

To take a Web application offline before deployment

Without speaking specifically to your points about downtime, the points I would interject are:

1. If your systems group is any good, they'll be able to create a Windows Shell Script or something to ease deployment in a manner that would bring servers down, update files, then brings them back up.

I can't help but wonder if when they say "it would take too long" they're assuming a manual process, but a competent systems group would look at the problems in terms of automating it to a point where they simply have to click a button.

Sure, a few servers will have limited downtime, but that brings me to the next point.

2. You have a five-server farm, and I presume it's prepared for failover. That's one of the advantages of having such hardware behind the system!

Just have them run a script during low-traffic and make sure they design the update such that different machines in the farm aren't running different versions of the app at the same time.

At any rate, this is not a development issue. It has nothing to do with your code and everything to do with update management. The responsibility of deploying without a hitch should fall squarely on the shoulders of a systems team, though you may be so gracious as to offer your help.

 

[jojo11]

Thanks so much for this feedback. I have been using some references to it in our conversations.

Does anyone have any other articles that talk about automated processes for taking a server out of a farm, udpating and putting back in the farm?

I realize that it would be different for defferent system setups, however I just need to see that other enterprise size web sites are doing this as well.

-------------------------------------------
Ummm, we have a bit of a problem here....