Open Stack

H Giacomo,

The patch set that you linked is 4 years old, so I don't remember what
the issue was with implementing the LB poll in the Senlin health
check.  I don't see a problem with implementing that feature so feel
free to check the contribution guide and submit some code.  A few
things to think about are:

- Would the LB_STATUS_POLLING require the user to use the Senlin Load
Balancer as well?
- How would Senlin know which health monitor to check for a given Senlin node?

A common use case for Senlin is autoscaling, so nodes will be added or
deleted on the fly.  Most likely that requires that the
LB_STATUS_POLLING needs to be tied to the Senlin LB policy that is
responsible for adding/removing LB pools and LB health monitors.  If
we go down that route, it is easy to retrieve the LB health monitor ID
because that is stored in the Senlin node properties.

Anyways, just some things to think about.

Duc

On Fri, May 14, 2021 at 2:51 AM Giacomo Lanciano
<giacomo.lanciano at sns.it> wrote:
>
> Hi folks,
>
> I'd like to know what is the status of making the LB_STATUS_POLLING
> detection mode available for the Senlin Health Policy [1]. According to
> the docs and this patch [2], the implementation of this feature is
> blocked by some issue on LBaaS/Octavia side, but I could not find any
> details on what this issue really is.
>
> As the docs state, it would be really useful to have this detection mode
> available, as it is much more reliable than the others at evaluating the
> status of an application. If I can be of any help, I would be willing to
> contribute.
>
> Thanks in advance.
>
> Kind regards.
>
> Giacomo
>
> [1]
> https://docs.openstack.org/senlin/latest/contributor/policies/health_v1.html#failure-detection
> [2] https://review.opendev.org/c/openstack/senlin/+/423012
>
> --
> Giacomo Lanciano
> Ph.D. Student in Data Science
> Scuola Normale Superiore, Pisa, Italy
> https://www.linkedin.com/in/giacomolanciano
>
>