-
Notifications
You must be signed in to change notification settings - Fork 414
mesos-consul deosn't deregister services when a worker node is shut down #1835
Comments
In this situation, mesos-consul is trying to deregister the service that was running on worker-001. In order to do that, however, it needs to connect to the agent on worker-001 to deregister. Since the node (and the agent) is down it cannot deregister the service. Consul should recognize that the service on worker-001 is unavailable and should not return it when queried. |
This is current case for all consul services irrelevant where they are deployed, even k8 pods stay visible after the pods have been destroyed |
@ChrisAubuchon are you able to assist with this issue? |
I've seen this issue in the past. It seems the worker that owns the
services has to shut down the service, not any others. Bit of a consul bug
if you ask me.
…On Tue, Apr 25, 2017 at 11:18 AM, Travis Rowland ***@***.***> wrote:
@ChrisAubuchon <https://github.com/ChrisAubuchon> are you able to assist
with this issue?
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#1835 (comment)>, or mute
the thread
<https://github.com/notifications/unsubscribe-auth/AKErsiOwJIV2hXmwE9S7NJWumhJcZ3WHks5rzhzWgaJpZM4KAxDq>
.
--
sr. devops engineer
wavefront
|
mesos-consul does not interact with Kubernetes so @KaGeN101's issue is unrelated to mesos-consul. |
I have seen all the k8 pods on the only consul there is. Consul is also k8s default store hence the k8-consul fork. Every pod you create in k8 ends up in consul and just stays there even if you destroy the pod |
It is not just mesos-consul it is consul for both mesos and k8 |
Using mantl 1.2.0
I have a mantl cluster running mesos with several services running on worker nodes.
When I shut-down the worker-001 node, marathon automatically relaunches the services running on worker-001 on other worker nodes (e.g. worker-003).
However, in consul the service remains registered on worker-001 (as well as on the new worker-003).
shouldn't mesos-consul detect that the service is no longer running on worker-001 and call "consul deregister"?
The text was updated successfully, but these errors were encountered: