Post

4 followers Follow
2
Avatar

F5 Big-IP orchestration within composite package

We're using the F5 Big-IP loadbalance plug-in (v4.5.0) with a composite package containing orchestrators sequential-by-deployment-group-level-2 and parallel-by-deployment-group. It's a mixture of WebLogic deployments to multiple domains and also a bunch of file.File and file.Folder types for things like properties files, static content, etc.

The deployment plan that is generated contains 2 top level sequential sections (driven by deployment group level 2). Each sequential section contains additional sections that are in parallel (driven by deployment group).

The 2 sequential sections represent a SideA and SideB. We take SideA out of the F5, update it, then bring it back in. Repeat with SideB.

The problem that we are having is that we need the very first step of all of SideA to remove the servers we have tied to SideA from the F5. Then, once ALL of SideA is updated we'd need the very last step to return the servers tied to SideA to the F5. Then the deployment can move onto SideB and do the same thing.

The current behavior is that within the SideA top-level sequential section there are multiple parallel sections that each contain a bunch of steps. One of those parallel sections contains the steps to remove servers from the F5 at the start of the section, performs a bunch of work on those servers, then returns those servers back to the F5 at the end of that section. The issue here is that we need all the other parallel sections that are executing at the same level (1 level deep within the top-level sequential section) to all be finished before we return any servers in this section to the F5. In this case, it is very likely that the servers that were removed from the F5 will be returned prior to the rest of SideA finishing. This will cause traffic to flow through SideA before the rest of it is updated, causing potential impact to our users.

I'm not sure how we can achieve this. Any help would be GREATLY appreciated.

Josh Figler

Please sign in to leave a comment.

8 comments

0
Avatar

Hi Josh,

It is difficult to follow the scenario by reading the text. Is it possible to post the screen shot of the current expanded deployment steps and draw an arrow where it goes wrong. It would help me to visualize the problem.

Levent Tutar 0 votes
0
Avatar

Hi Levent,

I wanted to do that originally, and I have an image ready, but I don't see a way to attach an image to this thread... am I missing it?

Josh Figler 0 votes
0
Avatar

Hi Levent -- any word on getting an image posted? Or is there anything else I can do to provide more information on what it is that we're looking for?

Josh Figler 0 votes
0
Avatar

Hey Levent and company -- anything we can do to make sure this thread doesn't get dropped. We are really interested in figuring out a solution to this.

Josh Figler 0 votes
0
Avatar

Hi Josh,

just to set the expectation :) I do not work at the XebiaLabs. I try to help the others in my spare time. I will have a look at your question this weekend. It isn't something I can answer in a couple of minutes.

Levent Tutar 0 votes
0
Avatar

Hi Levent,

Well you had me fooled! Thanks for taking the time to do that -- I can think of a few times you've helped me out.

And no wonder ltutar@xebialabs bounced back when I sent you the picture that I can't post here ;)

Thanks again.

Josh Figler 0 votes