Docker swarm can be accessed only on nodes, where container is running

Question

I'm currently running docker swarm on 3 nodes. First I created network as

docker network create -d overlay xx_net

after that a service as

docker service create --network xxx_net --replicas 1 -p 12345:12345 --name nameofservice nameofimage:1

If I read correctly, this is routing mesh (=ok for me). But I can only access service on that node-ip, where container is running, even it should be available on every node ip's.

If I drain some node, container starts up on different node and then it's on available on new ip.

---

**more information added below here:

• I rebooted all servers - 3 workers, where on of them is manager
• after boot, all seems to work ok!
• I'm using rabbitmq-image from docker hub. Dockerfile is quite small: FROM rabbitmq:3-management Container has been started at worker 2
• I can connect to rabbitmq's management page from all workers: worker1-ip:15672, worker2-ip:15672, worker3-ip:15672, so I think all ports needed is open.
• about after 1 hour, rabbitmq-container has been moved from worker 2 to worker 3 - I do not know reason.
• after that I cannot anymore connect from worker1-ip:15672, worker2-ip:15672 but from worker3-ip:15672 all still works!
• I drained worker3 as docker node update --availability drain worker3
• container started at worker1.
• after that I can only connect from worker1-ip:15672, not anymore from worker2 or worker3

One test more: all docker services restarted on all workers, and all works again?! - let's wait a few hours...

Today's status: 2 of 3 nodes are working ok. On service log of manager:

Jul 12 07:53:32 dockerswarmmanager dockerd[7180]: time="2017-07-12T07:53:32.787953754Z" level=info msg="memberlist: Marking dockerswarmworker2-459b4229d652 as failed, suspect timeout reached"
Jul 12 07:53:39 dockerswarmmanager dockerd[7180]: time="2017-07-12T07:53:39.787783458Z" level=info msg="memberlist: Marking dockerswarmworker2-459b4229d652 as failed, suspect timeout reached"
Jul 12 07:55:27 dockerswarmmanager dockerd[7180]: time="2017-07-12T07:55:27.790564790Z" level=info msg="memberlist: Marking dockerswarmworker2-459b4229d652 as failed, suspect timeout reached"
Jul 12 07:55:41 dockerswarmmanager dockerd[7180]: time="2017-07-12T07:55:41.787974530Z" level=info msg="memberlist: Marking dockerswarmworker2-459b4229d652 as failed, suspect timeout reached"
Jul 12 07:56:33 dockerswarmmanager dockerd[7180]: time="2017-07-12T07:56:33.027525926Z" level=error msg="logs call failed" error="container not ready for logs: context canceled" module="node/agent/taskmanager" node.id=b6vnaouyci7b76ol1apq96zxx
Jul 12 07:56:33 dockerswarmmanager dockerd[7180]: time="2017-07-12T07:56:33.027668473Z" level=error msg="logs call failed" error="container not ready for logs: context canceled" module="node/agent/taskmanager" node.id=b6vnaouyci7b76ol1apq96zxx
Jul 12 08:13:22 dockerswarmmanager dockerd[7180]: time="2017-07-12T08:13:22.787796692Z" level=info msg="memberlist: Marking dockerswarmworker2-03ec8453a81f as failed, suspect timeout reached"
Jul 12 08:21:37 dockerswarmmanager dockerd[7180]: time="2017-07-12T08:21:37.788694522Z" level=info msg="memberlist: Marking dockerswarmworker2-03ec8453a81f as failed, suspect timeout reached"
Jul 12 08:24:01 dockerswarmmanager dockerd[7180]: time="2017-07-12T08:24:01.525570127Z" level=error msg="logs call failed" error="container not ready for logs: context canceled" module="node/agent/taskmanager" node.id=b6vnaouyci7b76ol1apq96zxx
Jul 12 08:24:01 dockerswarmmanager dockerd[7180]: time="2017-07-12T08:24:01.525713893Z" level=error msg="logs call failed" error="container not ready for logs: context canceled" module="node/agent/taskmanager" node.id=b6vnaouyci7b76ol1apq96zxx

and from worker's docker log:

Jul 12 08:20:47 dockerswarmworker2 dockerd[677]: time="2017-07-12T08:20:47.486202716Z" level=error msg="Bulk sync to node h999-99-999-185.scenegroup.fi-891b24339f8a timed out"
Jul 12 08:21:38 dockerswarmworker2 dockerd[677]: time="2017-07-12T08:21:38.288117026Z" level=warning msg="memberlist: Refuting a dead message (from: h999-99-999-185.scenegroup.fi-891b24339f8a)"
Jul 12 08:21:39 dockerswarmworker2 dockerd[677]: time="2017-07-12T08:21:39.404554761Z" level=warning msg="Neighbor entry already present for IP 10.255.0.3, mac 02:42:0a:ff:00:03"
Jul 12 08:21:39 dockerswarmworker2 dockerd[677]: time="2017-07-12T08:21:39.404588738Z" level=warning msg="Neighbor entry already present for IP 104.198.180.163, mac 02:42:0a:ff:00:03"
Jul 12 08:21:39 dockerswarmworker2 dockerd[677]: time="2017-07-12T08:21:39.404609273Z" level=warning msg="Neighbor entry already present for IP 10.255.0.6, mac 02:42:0a:ff:00:06"
Jul 12 08:21:39 dockerswarmworker2 dockerd[677]: time="2017-07-12T08:21:39.404622776Z" level=warning msg="Neighbor entry already present for IP 104.198.180.163, mac 02:42:0a:ff:00:06"
Jul 12 08:21:47 dockerswarmworker2 dockerd[677]: time="2017-07-12T08:21:47.486007317Z" level=error msg="Bulk sync to node h999-99-999-185.scenegroup.fi-891b24339f8a timed out"
Jul 12 08:22:47 dockerswarmworker2 dockerd[677]: time="2017-07-12T08:22:47.485821037Z" level=error msg="Bulk sync to node h999-99-999-185.scenegroup.fi-891b24339f8a timed out"
Jul 12 08:23:17 dockerswarmworker2 dockerd[677]: time="2017-07-12T08:23:17.630602898Z" level=error msg="Bulk sync to node h999-99-999-185.scenegroup.fi-891b24339f8a timed out"

And this one from working worker:

Jul 12 08:33:09 h999-99-999-185.scenegroup.fi dockerd[10330]: time="2017-07-12T08:33:09.219973777Z" level=warning msg="Neighbor entry already present for IP 10.0.0.3, mac xxxxx"
Jul 12 08:33:09 h999-99-999-185.scenegroup.fi dockerd[10330]: time="2017-07-12T08:33:09.220539013Z" level=warning msg="Neighbor entry already present for IP "managers ip here", mac xxxxxx"

I restarted docker on problematic worker and it started to work again. I'll be following...

** Today's results:

• 2 of workers available, one is not
• I didn't a thing
• after 4 hour "swarm alone", all seems to works again?!
• services has been moved from worker to other for any good reason, all results seems to be problem with communication.
• quite confusing.

Answer 1

Upgrade to docker 17.06

Ingress overlay networking was broken for a long time until about 17.06-rc3