[datadog] add examples for DC/OS 1.10

[hkaj]

Hi @pleia2 Thanks for the thorough review! I addressed your feedback and fixed an issue in the auto discovery template that prevented this example from working. I followed it step by step and it should work now. Let me know if you see anything missing.

[pleia2]

This works perfectly for 1.10! And thank you for the improvements on the DC/OS side to make sure users are able to log into the nodes.

As for 1.11, I ran into a snag. The datadog agent configuration doesn't include a section for volumes in 1.11 like it does in 1.10, so I can't mount the required auto_conf.yaml. Are you seeing something different?

[hkaj]

Right, the menu for volume doesn't show up in 1.11, I don't know why. Passing auto_conf.yaml is not required, but to do it I ended up creating a service from scratch and not using the package from universe. Do you know why this menu isn't available in 1.11? Happy to write a PR to fix that if it's something we control on the package side. Otherwise I'll update this PR to reflect this limitation.

[pleia2]

The problem may be that we don't expose the Marathon definition details through the UI for 1.11 (I have questions out to our UI team about it, it's causing stress for tasks elsewhere too) and that may be what we're bumping into here. @ryadav88 do you have any thoughts here?

[hkaj]

@pleia2 should I update the example to reflect this limitation or wait for an update? What do you think of:

-You can do that easily in the DC/OS web UI by adding a new volume to the Datadog service. From the Services list, select Datadog, and then click the Edit button to modify the service parameters. Under "Volumes," add the following volume:
+To pass custom volumes to the Datadog service you will need to create the service from scratch as Universe packages don't allow changing mounted volumes. In this new service, under "Volumes" add the following volume on top of the default ones (as shown on the image below):

 * Container path: `/conf.d`
 * Host path: `/opt/datadog-agent-conf.d`
@@ -145,7 +145,7 @@ You can do that easily in the DC/OS web UI by adding a new volume to the Datadog

 ![Mounting a config volume in the Datadog Agent container](img/dcos-volume.png)

-Deploy the changes to the Datadog service. You can then verify that the configuration is correct by running the Datadog `status` command on your agent node. For MySQL the command and output should look something like the snippet below:
+Deploy the custom Datadog service. You can then verify that the configuration is correct by running the Datadog `status` command on your agent node. For MySQL the command and output should look something like the snippet below:

[pleia2]

Thanks @hkaj. I think your proposed text will work fine for now, though I will make a note to cycle back to this because it's obviously not the best solution. Are you affiliated with DataDog at all, or should I reach out to them?

[hkaj]

@pleia2 I'm an eng at Datadog :) If this PR is good for you, it's good for us.

[pleia2]

Ok great, go for it.

[hkaj]

Thanks! I don't have write access to this repo though. @ryadav88 can you help getting this merged please?

[pleia2]

I can merge it, you'll just want to update your PR with the changes you proposed.

[hkaj]

Ha my bad, i thought I had pushed them. Fixed.

[pleia2]

Hate to do this to you, but I made a suggestion inline to use the Marathon UI to adjust the volumes instead, since this appears to be the easiest way to do it in 1.11.

There's a public ticket tracking details about this 1.11 change that doesn't allow editing things like this https://jira.mesosphere.com/browse/DCOS_OSS-3679

[pleia2]

Thanks for this update! Can you also update datadog/README.md to add 1.10 and adjust the PR title since 1.11 isn't included in this?

[hkaj]

Hey @pleia2 I'm not sure how to do that (I'm using a slightly modified version of this cloudformation template to deploy a DC/OS cluster, and don't know how to Marathon from there), and from the docs it looks like the UI is deprecated anyway. So I removed the example for 1.11, and will add it again in a separate PR when there is a better solution.

Are we good with the 1.10 example?

[hkaj]

Hadn't seen your latest message, fixed.

[pleia2]

The Marathon UI is being deprecated, but since right now it's the only way we have to make this change through a UI, we're not sure where things are going to land with this. Anyway, thank you for this update for 1.10, merging!