If you're using the Helios vagrant image (as described below), you shouldn't need to set any environment variables to run the integration tests.
Otherwise,
- Set
ETCD_SERVER
to location of your etcd server, e.g.http://127.0.0.1:4001
. - Set
SKYDNS_SERVER
to the address of your SkyDNS server. - If you're running SkyDNS on a port that isn't 53, set
SKYDNS_PORT
to that port.
To run the integration tests, run maven verify
.
Regular unit tests shouldn't require anything to run.
To run the helios agent or master with this registrar:
- Set the
--service-registrar-plugin
option to the location of the jar file. - Set the
--service-registry
option to the location of the etcd server you want to talk to, e.g.http://192.168.33.10:4001
. - Set the
--domain
option to which domain you want things to register themselves in.
If you're using the Helios Vagrant image, you can run the integration tests here against that and they should just work.
If not, you'll need to do something akin to this, adjusting $GOPATH
and the 192.168.33.10
address bits below:
# install golang and the VC systems it needs to `get` things
sudo apt-get -y --force-yes install git-core
sudo apt-get -y --force-yes install mercurial
sudo apt-get -y --force-yes install golang
# create a place to put stuff
export GOPATH=/vagrant/gopath
mkdir $GOPATH
# get, build and install SkyDNS
git clone [email protected]:skynetservices/skydns.git
cd skydns
go get -d -v ./... && go build -v ./...
sudo cp skydns /usr/bin
cd ..
# get, build and install etcd
git clone https://github.com/coreos/etcd
cd etcd
./build
sudo cp bin/etcd /usr/bin
cd ..
# start etcd and skydns
etcd & disown
sudo skydns -addr 192.168.33.10:53 -machines="127.0.0.1:4001" -nameservers="8.8.8.8:53,8.8.4.4:53"
These above instructions should work on debian or ubuntu machines generally. The apt-get
lines should be the only thing to change on non-ubuntu machines.
By default it will register your jobs with the format
service.protocol.domain, but this is overrideable via the
REGISTRAR_HOST_FORMAT
environment variable. It is a
org.apache.commons.lang3.text.StrSubstitutor
format string. So the
actual default format is specified in the code as
${service}.${protocol}.${domain}
. You can change this as you deem
necessary.
To run findbugs on this codebase, do
mvn clean compile site
. This will build helios and then run an analysis,
emitting reports in target/site/findbugs.html
.
To silence an irrelevant warning, add a filter match along with a justification
in findbugs-exclude.xml
.
A simple job definition to try things out with is:
helios create -p echo=4711 -r echo/text=echo echojob:1 busybox -- nc -p 4711 -lle cat
It's a simple echo server, which you then can deploy and then you should be able to look it up in DNS (assuming your DNS infrastructure is set up properly) following the SRV record format.
Something I've found useful in debugging is dumping the contents of SkyDNS, or actually etcd.
curl -L http://youretcdhost:4001/v2/keys/skydns?recursive=true | jq -C . | less -r
Since crud may accumulate in the SkyDNS subtree of etcd, especially if using the Helios
Testing Framework, you may want to install SkyGC to
clean up the leftovers periodically. But etcd can go on for a loooonng time before the
build up of trash will cause problems -- so much so that this was written because the
dumps (from the curl
example above) got so long that it was a pain to sift through, not
because etcd had a problem.