- fixes startCluster's --wait/--timeout functionality
--> now doesn't check all the logs (e.g. from previous run)
----> no more false positives when checking for a specifc log message
only in the current run
- package cluster: use context.Context as first function param in every
exported function
- package cmd: pass cmd.Context() to calls to package cluster exported
functions
- masterlb is now an extra part of the cluster spec
- ports can now be attached to masterlb by using the 'loadbalancer'
nodefilter
- all ports exposed on the masterlb will be proxied to all master nodes
(on the same port)
- we now use the full containerJSON details when getting a node
- we now use as many details as possible to copy k3d settings from an
existing node when adding a new node to a running cluster
- if no node with the target role exists in the cluster, just choose any
other (non-proxy) node instead
- to make this work, we also need to add the k3s url label to every node
- refactoring in pkg/cluster/kubeconfig
- new functions to delete a cluster from kubeconfig
- call RemoveClusterFromDefaultKubeConfig after deleting cluster
- enables wait-for-master
- calls GetAndWriteKubeConfig after successful cluster creation to
update the default kubeconfig with the new cluster's details
- does NOT automatically switch the current-context
- outputs a different line saying, that you can switch context now
- imageVolume couldn't be determined before in some cases, because it
tried to extract the name from labels on the masterlb which doesn't have
them.
- now we're only trying get the label from master/worker nodes and this
also adds a failover solution
- also, we ignore non-worker/master nodes upon image import
Before this change, we simply did a search/replace on the
stringified kubeconfig blob.
Now we're parsing it into a kubeconfig struct and modify the fields
directly in a more controlled manner.
Here's what we change:
- server URL: based on the chosen APIHost and APIPort
- cluster name: default -> k3d-CLUSTERNAME
- user name: default -> admin@k3d-CLUSTERNAME
- context name: default -> admin@k3d-CLUSTERNAME
With the updated cobra depencendy, we're now passing a context
from the cmd to the called functions.
When creating a cluster, one can pass a Duration to the --timeout
flag, which will create a new context with a timeout.
In the two blocking functions, where we're waiting for the master nodes
(initializing master nodes and "normal" master nodes), we're now
checking for the context cancellation as well, which may be caused
by the timeout.
Up to now, we exposed ports on single master nodes, which is quite
inconvenient on user side and troublesome on development side.
Now, we're creating a proxy container which exposes a single port
and proxies traffic to all master nodes.
Currently, this only works with 'k3d create cluster' and won't
update the proxy when using 'k3d create node --role master'.