Hello, have you tried to use tcp check rather than http check ?

Il giorno mar 6 apr 2021 alle ore 17:38 Adekunbi Adewojo aadewojo@gmail.com ha scritto:

Thank you very much for your detailed response. I checked my previous loadbalancer implementation, the members operating status showed active. However, when I checked the access log of one of the load balancer member it showed this "06/Apr/2021:06:25:05 +0000] "GET /healthcheck HTTP/1.0" 404 118".

I then deleted the load balancer and recreated it. I realised that before adding a listener or any other thing, the load balancer wasn't showing an "Online" status as suggested by the cookbook. I also ran the stat command and everything was zero.

I see that you mentioned neutron, I do not have admin access, I might have to go back to the admin. But from what I said, do you still think it is a neutron issue?

Thanks.

On Tue, Apr 6, 2021 at 4:12 PM Michael Johnson johnsomor@gmail.com wrote:

...

Hi Adekunbi,

It sounds like the backend servers (web servers?) are not passing the health check or are otherwise unreachable. Provisioning status of Active shows that Octavia was able to create and provision the load balancer without error.

Let's look at a few things:

1. Check if the that load balancer has statistics for the your

connections to the VIP:

openstack loadbalancer stats show <lb name or ID>

If these are all zeros, your deployment of Octavia is not working correctly. Most likely the lb-mgmt-net is not passing the required traffic. Please debug in neutron.

Assuming you see a value greater than zero in the "total_connections" column, your deployment is working as expected.

2. Check your health monitor configuration and load balancer status:

   openstack loadbalancer status show <lb name or ID>

Check the "operating status" of all of the objects in your load balancer. As a refresher, operating status is the observed status of the object, so do we see the backend member as ONLINE, etc.

openstack loadbalancer member show <pool ID or name> <member ID or

name>

Also check that the member is configured with the correct subnet that can reach the backend member server. If a subnet was not specified, it will use the VIP subnet to attempt to reach the members.

If the members are in operating status ERROR, this means that the load balancer sees that server as failed. Check your health monitor configuration (If you have one) to make sure it is connecting to the correct IPs and ports and the expected response is correct for your application.

openstack loadbalancer healthmonitor show <health monitor ID>

Also, check that the members have security groups or other firewall runs set appropriately to allow the load balancer to access it.

Michael

On Fri, Apr 2, 2021 at 8:36 AM Adekunbi Adewojo aadewojo@gmail.com wrote:

...

Hi there,

I recently deployed a load balancer on our openstack private cloud. I

used this manual - https://docs.openstack.org/octavia/latest/user/guides/basic-cookbook.html

...

to create the load balancer. However, after creating and trying to

access it, it returns an error message saying "No server is available to handle this request". Also on the dashboard, "Operating status" shows offline but "provisioning status" shows active. I have two web applications as members of the load balancer and I can individually access those web applications.

...

Could someone please point me in the right direction?

Thanks.