Igor is a service that provides a single point of integration with Continuous Integration (CI) and Source Control Management (SCM) services for Spinnaker.
Igor runs a number of pollers that all share the same common architecture. At a high level, they all:
- periodically get a list of items from an external resource (e.g. builds on a Jenkins master)
- compare that list against their own persisted cache of items (the difference is called delta size)
- send an echo event for each new item
- cache the new list of items
Features:
- health: igor has a
HealthIndicator
that reportsDown
if no pollers are running or if they have not had a successful polling cycle in a long time - locking: pollers can optionally acquire a distributed lock in the storage system before attempting to complete a polling cycle. This makes it possible to run igor in a high-availability configuration and scale it horizontally.
- safeguards: abnormally large delta sizes can indicate a problem (e.g. lost or corrupt cache data) and cause downstream issues. If a polling cycle results in a delta size above the threshold, the new items will not be cached and events will not be submitted to echo to prevent a trigger storm. Manual action will be needed to resolve this, such as using the fast-forward admin endpoint:
/admin/pollers/fastforward/{monitorName}[?partition={partition}]
. Fast-forwarding means that all pending cache state will be polled and saved, but will not send echo notifications.
Relevant properties:
Property | Default value | Description |
---|---|---|
spinnaker.build.pollInterval |
60 |
Interval in seconds between polling cycles |
spinnaker.pollingSafeguard.itemUpperThreshold |
1000 |
Defines the upper threshold for number of new items before a cache update cycle will be rejected |
locking.enabled |
false |
Enables distributed locking so that igor can run on multiple nodes without interference |
Relevant metrics:
Metric | Type | Description |
---|---|---|
pollingMonitor.newItems |
gauge | represents the number of new items cached by a given monitor during a polling cycle |
pollingMonitor.itemsOverThreshold |
gauge | 0 if deltaSize < threshold, deltaSize otherwise |
pollingMonitor.pollTiming |
timer | published for every polling cycle with the duration it took to complete |
pollingMonitor.failed |
counter | an error counter indicating a failed polling cycle |
All these metrics can be grouped by a monitor
tag (e.g. DockerMonitor
, JenkinsMonitor
...) to track down issues.
The following storage backends are supported:
- Redis
Relevant properties:
redis:
enabled: true
connection: redis://host:port
The following SCM services are supported:
- Bitbucket
- Github
- Gitlab
- Stash
Commit
controller classes expose APIs to retrieve lists of commits, such as /github/{{projectKey}}/{{repositorySlug}}/compareCommits?from={{fromHash}}&to={{toHash}}
At the moment, igor only exposes read APIs, there are no pollers and no triggers involving SCM services directly.
Relevant properties:
github:
baseUrl: "https://api.github.com"
accessToken: '<your github token>'
commitDisplayLength: 8
stash:
baseUrl: "<stash url>"
username: '<stash username>'
password: '<stash password>'
bitbucket:
baseUrl: "https://api.bitbucket.org"
username: '<bitbucket username>'
password: '<bitbucket password>'
commitDisplayLength: 7
gitlab:
baseUrl: "https://gitlab.com"
privateToken: '<your gitlab token>'
commitDisplayLength: 8
The following CI services are supported:
- Artifactory
- Nexus
- Concourse
- Gitlab CI
- Google Cloud Build (GCB)
- Jenkins
- Travis
- Wercker
For each of these services, a poller can be enabled (e.g. with jenkins.enabled
) that will start monitoring new builds/pipelines/artifacts, caching them and submitting events to echo, thus supporting pipeline triggers. GCB is a bit different in that it doesn't poll and requires setting up pubsub subscriptions.
The BuildController
class also exposes APIs for services that support them such as:
- getting build status
- listing builds/jobs on a master
- listing queued builds
- starting and stopping builds/jobs
These APIs are used to provide artifact information for bake stages.
In your configuration block (either in igor.yml, igor-local.yml, spinnaker.yml or spinnaker-local.yml), you can define multiple masters blocks by using the list format.
You can obtain a Jenkins API token by navigating to http://your.jenkins.server/me/configure
(where me
is your username).
jenkins:
enabled: true
masters:
-
address: "https://spinnaker.cloudbees.com/"
name: cloudbees
password: f5e182594586b86687319aa5780ebcc5
username: spinnakeruser
-
address: "http://hostedjenkins.amazon.com"
name: bluespar
password: de4f277c81fb2b7033065509ddf31cd3
username: spindoctor
In your configuration block (either in igor.yml, igor-local.yml, spinnaker.yml or spinnaker-local.yml), you can define multiple masters blocks by using the list format.
To authenticate with Travis you use a "Personal access token" on a git user with permissions read:org, repo, user
. This is added in settings -> Personal access tokens
on github/github-enterprise.
travis:
enabled: true
# Travis names are prefixed with travis- inside igor.
masters:
- name: ci # This will show as travis-ci inside spinnaker.
baseUrl: https://travis-ci.com
address: https://api.travis-ci.com
githubToken: 6a7729bdba8c4f9abc58b175213d83f072d1d832
regexes:
- /Upload https?:\/\/.+\/(.+\.(deb|rpm))/
When parsing artifact information from Travis builds, igor uses a default regex
that will match on output from the jfrog rt
/art
CLI tool. Different regexes than the
default may be configured using the regexes
list.
Clouddriver can be configured to poll your registries. When that is the case, igor can then create a poller that will list the registries indexed by clouddriver, check each one for new images and submit events to echo (hence allowing Docker triggers)
Relevant properties:
dockerRegistry.enabled
- requires
services.clouddriver.baseUrl
to be configured
Igor requires redis server to be up and running.
Start igor via ./gradlew bootRun
. Or by following the instructions using the Spinnaker installation scripts.
To start the JVM in debug mode, set the Java system property DEBUG=true
:
./gradlew -DDEBUG=true
The JVM will then listen for a debugger to be attached on port 8188. The JVM will not wait for the debugger to be attached before starting igor; the relevant JVM arguments can be seen and modified as needed in build.gradle
.