Helping new users with Cloud Foundry

[Quintessence Anx]

Hello all!

If you were to explain the Health Manager to someone so they can set up and run Cloud Foundry in Production, how would you do it?

Thank you!

Quinn

[James Bayer]

the health manager provides automated health recovery for applications. it is responsible for converging the "desired" state of the world as reflected by API calls and the persistent model with the "actual" state of the world from the CF system application monitoring. when there is a different, the health manager will tell the cloud controller that it noticed a discrepancy, which should then be corrected if it is possible (enough memory, etc).

--
You received this message because you are subscribed to the Google Groups "Cloud Foundry Developers" group.
To view this discussion on the web visit https://groups.google.com/a/cloudfoundry.org/d/msgid/vcap-dev/CAFyae%2BaKe8%2B74NTpCGL2s1nD%3D1DkwzTK1FNkguL9i-KPsR1Utg%40mail.gmail.com.

To unsubscribe from this group and stop receiving emails from it, send an email to vcap-dev+u...@cloudfoundry.org.

--

Thank you,

James Bayer

[Quintessence]

I tried to create a diagram that represents that flow of information ("picture is worth 1000 words" and what not) and I was wondering if you (open to all, of course) could give me your thoughts/feedback?

http://www.gliffy.com/go/publish/image/6567409/L.png

A couple of things that I'm not really clear on for the HM:

* The HM checks the ideal state of the system via the Cloud Controller (CC). When there is an app issue, too many/few/etc., the HM sends start stop messages over NATS. It was my understanding that the HM also checked the health of other components of the system via the CC. What happens if there is a problem with another component? e.g. a warden container, the blob store, the message bus... Or does the HM only monitor the applications (via the DEAs)?

* On a high level, what is the common context that users (developers or CF ops) would interact with the HM directly? 

* On a slightly more detailed-but-still-pretty-high level: how do you troubleshoot the HM itself if there is an issue?

[Ken Krueger]

qanx,

If it helps, here is some free training material on this very topic.  See this video, at approximately the 13:21 point in the video:https://pivotallms.biglms.com/courses/PivotalAcademy/1111/1111/courseware/83e42c47faa44f958bbddbc0a96f7b7a/bfdd894bd27048be8c319922b6091f32/  Note that Diego will change how a lot of this works (no health monitor!).

On the second part of your question, you are correct, the HealthMonitor detects apps that are not running, but is not monitoring or managing the other big processes like the blob store, message bus, cloud controller, etc.  Instead, this is where the BOSH Health Manager comes to the rescue by making sure sure that these VMs/processes are up and running.  This is part of the "4 levels of HA" that you've probably heard about.

Hope that helps,

k

To view this discussion on the web visit https://groups.google.com/a/cloudfoundry.org/d/msgid/vcap-dev/c3068e7f-1de3-4063-a05b-45e39d53ee04%40cloudfoundry.org.

To unsubscribe from this group and stop receiving emails from it, send an email to vcap-dev+u...@cloudfoundry.org.

--

Ken Krueger  

Manager, Global Education Delivery

407 256 9737 Mobile

kenkrueger65 Skype

Education questions?  educ...@pivotal.io

[Quintessence Anx]

Thank you for the resource Ken!

I have a question: the presentation linked states that the Health Manager does not communicate with the DEAs directly - but I thought that the Health Manager did communicate directly over NATS w/ START/STOP messages (per http://blog.cloudfoundry.org/2014/02/22/hm9000-ready-for-launch/). Or is this a difference between the HM9000 and the previous Health Manager?

Thanks!

--
You received this message because you are subscribed to a topic in the Google Groups "Cloud Foundry Developers" group.
To view this discussion on the web visit https://groups.google.com/a/cloudfoundry.org/d/msgid/vcap-dev/CA%2BYXpzmwV%3Dw11-d_VjRWU%2BzrLrFDEsP2m9%2BJA7xMukwkum%3DtKw%40mail.gmail.com.

[James Bayer]

the health manager sends information about app desired/actual state mismatching to the cloud controller. the cloud controller is responsible for taking action on things that the health manager notices.

To view this discussion on the web visit https://groups.google.com/a/cloudfoundry.org/d/msgid/vcap-dev/CAFyae%2BbWkmWkNMXOwgNBhZ40kCAbZPBQ22_25acGF0xRQF3cRw%40mail.gmail.com.

To unsubscribe from this group and stop receiving emails from it, send an email to vcap-dev+u...@cloudfoundry.org.

[Quintessence Anx]

Thank you! Just so I'm 100% clear: the HM9000 sends the START/STOP messages to the CC and NOT directly to the DEAs, correct?

Thanks again!

To view this discussion on the web visit https://groups.google.com/a/cloudfoundry.org/d/msgid/vcap-dev/CAB%3Dt-sV0OCN2Ps9JWROYwKfSKhowpBW1GFZhqkcXiPa-Jy%2BxXg%40mail.gmail.com.

[James Bayer]

correct.

To view this discussion on the web visit https://groups.google.com/a/cloudfoundry.org/d/msgid/vcap-dev/CAFyae%2BbVQ7Owijcg30SbX1gyQmTosyF5pfA-s%2B1XvBkKDW14MQ%40mail.gmail.com.

To unsubscribe from this group and stop receiving emails from it, send an email to vcap-dev+u...@cloudfoundry.org.

[Quintessence]

Thank you for clearing that up for me :)

I made a diagram to help me visualize what we've discussed about the Health Manager. I put it here:

http://www.gliffy.com/go/publish/image/6567409/L.png

(If the "L" size is too large you can change it to "M" or "S" for smaller views of the same image.)

Is this accurate?

[James Bayer]

that looks correct to me, but i haven't reviewed all of the code paths.

To view this discussion on the web visit https://groups.google.com/a/cloudfoundry.org/d/msgid/vcap-dev/e179b171-f014-457f-b11c-dc9a9c7b17de%40cloudfoundry.org.

To unsubscribe from this group and stop receiving emails from it, send an email to vcap-dev+u...@cloudfoundry.org.

[Cornelia Davis]

The health manager does not communicate with the DEAs to issue start or stop commands, that is the job of the cloud controller. If the HM detects a discrepancy between the actual and desired states it advises the cloud controller of that difference and the cloud controller will signal the DEAs to make the necessary changes.

Sent from my iPhone

To view this discussion on the web visit https://groups.google.com/a/cloudfoundry.org/d/msgid/vcap-dev/CAFyae%2BbWkmWkNMXOwgNBhZ40kCAbZPBQ22_25acGF0xRQF3cRw%40mail.gmail.com.