high availability - early-docker.service

[det...@cisco.com]

Hi all, i'm running now my cluster properly.

member 18b68d8c1abcecd8 is healthy: got healthy result from http://10.50.204.65:2379
member 75bec168ef726ab4 is healthy: got healthy result from http://10.50.204.84:2379
member d6fc4d677c0e8c50 is healthy: got healthy result from http://10.50.204.68:2379
cluster is healthy

Moving to kubernetes configuration, i did configure my early-docker.service as following (default configuration).

[Unit]
Description=Early Docker Application Container Engine
Documentation=http://docs.docker.com
After=early-docker.socket
Requires=early-docker.socket
[Service]
Environment="DOCKER_CGROUPS=--exec-opt native.cgroupdriver=systemd"
Environment=TMPDIR=/var/tmp
MountFlags=slave
LimitNOFILE=1048576
LimitNPROC=1048576
ExecStart=/usr/lib/coreos/dockerd daemon --host=fd:// --bridge=none --iptables=false --ip-masq=false --graph=/var/lib/early-docker --pidfile=/var/run/early-docker.pid $DOCKER_OPTS $DOCKER_CGROUPS
[Install]
WantedBy=early-docker.target
[X-Fleet]
Global=true

all necessary SW packages are available on each node :

REPOSITORY                           TAG                 IMAGE ID            CREATED             SIZE
gcr.io/google_containers/hyperkube   v1.1.4              59b39a4a07c5        4 months ago        253.3 MB
quay.io/coreos/flannel               0.5.3               95ae7a5bca26        9 months ago        21.41 MB
kubernetes/pause                     latest              f9d5de079539        23 months ago       239.8 kB

i then did a fleetctl start early-docker.service on each node (probably the mistake as it needs to run on one node only) and did check

sudo fleetctl list-units
UNIT            MACHINE                ACTIVE    SUB
early-docker.service    88731e6a.../10.50.204.65    active    running

My issue is that i only see one node available whereas i should have 3.

Would you have any idea ? maybe release the etcd2 DB should fix my issue ?

Thanks BR Olivier


[Seán C. McCord]

CoreOS comes with early-docker.service; you shouldn't have had to construct it.  It is generally for initializing flannel, and as such, you would run it on all nodes.  Moreover, you would not run it under fleet.

What is it you are trying to accomplish?

--
You received this message because you are subscribed to the Google Groups "CoreOS User" group.
To unsubscribe from this group and stop receiving emails from it, send an email to coreos-user...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

Seán C McCord

CyCore Systems, Inc

+1 888 240 0308