It is extremely unfortunate that the Helm team has provided virtually zero direction about how to go through the migration from 2 to 3. (I’m less impressed with Helm as a team and company every day. The evidence keeps piling up.)
Something as basic as detecting which version is installed you would think would have been the fucking first thing they documented. (Perhaps they did, but after an hour I could find nothing.) So here’s the best researched solution I could come up with in a few hours.
First we have to get all the user namespaces:
names() {
IFS=
while read -r line; do
name="${line%% *}"
test "$name" = NAME && continue
echo "$name"
done < <(kubectl get ns -l "nstype=user")
}
Then we have to determine if Tiller (a Helm 2 only thing) is in that namespace specifically:
tiller_exists() {
kubectl get deploy -n "$1" | grep tiller-deploy &>/dev/null
}
That will produce a 0 if tiller-deploy is included in the deployments,
which is true for every new user under Helm 2 (as far as I can tell).
We still need to check if Tiller is actively running in that namespace at the moment in order to run the upgrade script, otherwise it will just fail. It can be installed and just not running:
tiller_is_running() {
helm2 list "--tiller-namespace=$1" &>/dev/null
}
Normally this produces something like the following if not running:
Error: could not find tiller
Again, Tiller is there, just not running at the moment, which is why
the first tiller_exists is needed.