GluesyncPlatform reference

The GluesyncPlatform resource (API group gluesync.molo17.com/v1alpha1, short name gsp) describes one Gluesync install: the Core Hub version it should run, who may update it, and how the Gluesync Kubernetes Operator (GKO) checks an update. GKO serves one GluesyncPlatform per namespace, named gluesync by the kit and by the chart.

Example

apiVersion: gluesync.molo17.com/v1alpha1
kind: GluesyncPlatform
metadata:
  name: gluesync
  namespace: gluesync-dev
spec:
  version: "2.3.2.2"
  modules:
    chronos: "0.3.3"
  updatePolicy: ui
  stability:
    windowSeconds: 20
    maxWaitSeconds: 180
  maintenance:
    drainTimeoutSeconds: 600
  coreHub:
    url: https://gluesync-gluesync-core-hub:1717
    insecureSkipVerify: true
    moduleAuth:
      licenseSecretRef:
        name: gluesync-gluesync-shared
        key: gs-license.dat
      tag: conductor
  checkIntervalMinutes: 360

Edit it with kubectl edit gsp gluesync, kubectl patch, or in Git. When the kit or the chart creates it, the matching Helm values are listed in the last column of the tables below: operator.* for the kit, platform.* for the GKO chart.

Versions

Field Default Description Helm value

Field	Default	Description	Helm value
`version`	empty	Core Hub version to run, the image tag without the registry, for example `2.3.2.2`. Empty keeps what runs. When it differs from the running version, GKO starts an update.	kit: `gluesyncCoreHub.tag`; chart: `platform.version`
`modules`	empty	Module versions pinned by image name, with or without its `gluesync-` prefix: `chronos` and `gluesync-chronos` are the same module. A pinned module stays on its pin. An unpinned module moves to the latest version on the release channel each time Core Hub is updated. Changing only a pin updates that module, without maintenance mode.	chart: `platform.modules`
`releaseChannel`	empty	`ga`, `beta` or `alpha`. Leave it empty: GKO then reads the channel set in Core Hub, so the UI and GKO always agree. Set it only for an offline install.	chart: `platform.releaseChannel`
`requestID`	empty	Set to a new value by every update request from the UI. A new value retries an update to the same version that failed or was refused. Do not set it by hand, except to retry, see Retry an update.	none

version

empty

Core Hub version to run, the image tag without the registry, for example 2.3.2.2. Empty keeps what runs. When it differs from the running version, GKO starts an update.

kit: gluesyncCoreHub.tag; chart: platform.version

modules

empty

Module versions pinned by image name, with or without its gluesync- prefix: chronos and gluesync-chronos are the same module. A pinned module stays on its pin. An unpinned module moves to the latest version on the release channel each time Core Hub is updated. Changing only a pin updates that module, without maintenance mode.

chart: platform.modules

releaseChannel

empty

ga, beta or alpha. Leave it empty: GKO then reads the channel set in Core Hub, so the UI and GKO always agree. Set it only for an offline install.

chart: platform.releaseChannel

requestID

empty

Set to a new value by every update request from the UI. A new value retries an update to the same version that failed or was refused. Do not set it by hand, except to retry, see Retry an update.

none

When you start an update from the UI, GKO itself writes version and, for a module you update on its own, a pin in modules.

Who may update

Field Default Description Helm value

Field	Default	Description	Helm value
`updatePolicy`	`ui`	`ui`: updates start from Platform updates in the Core Hub UI, or from a change to `version`. `gitops`: the UI cannot start updates. It answers that updates of this install are managed in Git. Change `version` in Git instead, see Update from Git and Argo CD.	kit: `operator.updatePolicy`; chart: `platform.updatePolicy`

updatePolicy

ui

ui: updates start from Platform updates in the Core Hub UI, or from a change to version.
gitops: the UI cannot start updates. It answers that updates of this install are managed in Git. Change version in Git instead, see Update from Git and Argo CD.

kit: operator.updatePolicy; chart: platform.updatePolicy

The UI can start an update only for users with the SUPER_ADMIN or MANAGER role. GKO checks the user’s role with Core Hub on each request.

Stability check

Field Default Description Helm value

Field	Default	Description	Helm value
`stability.windowSeconds`	`20`	Seconds a component must stay Ready, without a container restart, to count as up. A restart during the window starts it again. `0` uses the default.	kit: `operator.stability.windowSeconds`; chart: `platform.stability.windowSeconds`
`stability.maxWaitSeconds`	`180`	Seconds from the image change until the component must have passed the window. The time spent pulling the image counts. Minimum `10`.	kit: `operator.stability.maxWaitSeconds`; chart: `platform.stability.maxWaitSeconds`

stability.windowSeconds

20

Seconds a component must stay Ready, without a container restart, to count as up. A restart during the window starts it again. 0 uses the default.

kit: operator.stability.windowSeconds; chart: platform.stability.windowSeconds

stability.maxWaitSeconds

180

Seconds from the image change until the component must have passed the window. The time spent pulling the image counts. Minimum 10.

kit: operator.stability.maxWaitSeconds; chart: platform.stability.maxWaitSeconds

These fields match Conductor’s UPDATE_VERIFY_SECONDS and UPDATE_VERIFY_MAX_WAIT_SECONDS. Raise maxWaitSeconds when Core Hub takes long to start, or when your registry is slow to serve a new image, for example a pull-through cache that has not cached it yet. See Stability check.

Maintenance mode

Field Default Description Helm value

Field	Default	Description	Helm value
`maintenance.drainTimeoutSeconds`	`600`	Seconds the running pipelines get to reach maintenance mode before Core Hub changes. When they do not make it in time, GKO resumes them and refuses the update, without changing any image. Minimum `10`.	kit: `operator.maintenance.drainTimeoutSeconds`; chart: `platform.maintenance.drainTimeoutSeconds`
`maintenance.disabled`	`false`	Skips maintenance mode: pipelines keep running while Core Hub restarts. Not recommended.	kit: `operator.maintenance.disabled`; chart: `platform.maintenance.disabled`

maintenance.drainTimeoutSeconds

600

Seconds the running pipelines get to reach maintenance mode before Core Hub changes. When they do not make it in time, GKO resumes them and refuses the update, without changing any image. Minimum 10.

kit: operator.maintenance.drainTimeoutSeconds; chart: platform.maintenance.drainTimeoutSeconds

maintenance.disabled

false

Skips maintenance mode: pipelines keep running while Core Hub restarts. Not recommended.

kit: operator.maintenance.disabled; chart: platform.maintenance.disabled

Core Hub access

Field Default Description Helm value

Field	Default	Description	Helm value
`coreHub.url`	`https://<core hub workload>.<namespace>.svc:1717`	URL of Core Hub’s REST API.	chart: `platform.coreHub.url`
`coreHub.insecureSkipVerify`	`false`	Accepts Core Hub’s self-signed certificate. The kit and the chart set `true`.	chart: `platform.coreHub.insecureSkipVerify`
`coreHub.moduleAuth.licenseSecretRef`		`name` and `key` of the Secret holding `gs-license.dat`. GKO connects to Core Hub as an external module with this licence. Recommended.	chart: `platform.coreHub.moduleAuth.licenseSecret`
`coreHub.moduleAuth.tag`	`conductor`	Tag GKO registers under. With `conductor`, Core Hub treats GKO as its Conductor module and forwards `/conductor/api` to it. Keep the default.	chart: `platform.coreHub.moduleAuth.tag`
`coreHub.moduleAuth.advertisePort`, `advertiseScheme`	`1717`, empty	Port and scheme (`http` or `https`) of the GKO API that Core Hub calls. The chart sets the scheme from `api.tlsSecretName`.	set by the chart
`coreHub.tokenSecretRef`		`name` and `key` (default `token`) of a Secret holding a Core Hub API token allowed to list pipelines and to enter and exit maintenance mode. Takes precedence over `moduleAuth`.	chart: `platform.coreHub.auth: token`, `platform.coreHub.tokenSecret`

coreHub.url

https://<core hub workload>.<namespace>.svc:1717

URL of Core Hub’s REST API.

chart: platform.coreHub.url

coreHub.insecureSkipVerify

false

Accepts Core Hub’s self-signed certificate. The kit and the chart set true.

chart: platform.coreHub.insecureSkipVerify

coreHub.moduleAuth.licenseSecretRef

name and key of the Secret holding gs-license.dat. GKO connects to Core Hub as an external module with this licence. Recommended.

chart: platform.coreHub.moduleAuth.licenseSecret

coreHub.moduleAuth.tag

conductor

Tag GKO registers under. With conductor, Core Hub treats GKO as its Conductor module and forwards /conductor/api to it. Keep the default.

chart: platform.coreHub.moduleAuth.tag

coreHub.moduleAuth.advertisePort, advertiseScheme

1717, empty

Port and scheme (http or https) of the GKO API that Core Hub calls. The chart sets the scheme from api.tlsSecretName.

set by the chart

coreHub.tokenSecretRef

name and key (default token) of a Secret holding a Core Hub API token allowed to list pipelines and to enter and exit maintenance mode. Takes precedence over moduleAuth.

chart: platform.coreHub.auth: token, platform.coreHub.tokenSecret

Images and catalog

Field Default Description Helm value

Field	Default	Description	Helm value
`imageRegistry`	empty	Registry and path that replace those of every image GKO sets, for example `registry.example.com/molo17`. The image name is kept: Core Hub moves to `registry.example.com/molo17/gluesync-core-hub:<version>`. Empty keeps each image’s own registry. Your mirror must hold every image GKO updates, Grafana and Prometheus included.	chart: `platform.imageRegistry`
`checkIntervalMinutes`	`360`	Minutes between two checks of the version catalog, which fill the Available column and the `UpdateAvailable` phase. Minimum `5`. To check at once, see Check for updates now.	none
`offline`	`false`	Stops every call to the version catalog. See Offline installs.	chart: `platform.offline`

imageRegistry

empty

Registry and path that replace those of every image GKO sets, for example registry.example.com/molo17. The image name is kept: Core Hub moves to registry.example.com/molo17/gluesync-core-hub:<version>. Empty keeps each image’s own registry. Your mirror must hold every image GKO updates, Grafana and Prometheus included.

chart: platform.imageRegistry

checkIntervalMinutes

360

Minutes between two checks of the version catalog, which fill the Available column and the UpdateAvailable phase. Minimum 5. To check at once, see Check for updates now.

none

offline

false

Stops every call to the version catalog. See Offline installs.

chart: platform.offline

GKO’s own update

GKO 0.2.0 and later update their own Deployment, see GKO’s own update. GKO’s version is pinned like a module, in modules, under gluesync-operator or operator. The other settings are Helm values, not fields of the GluesyncPlatform:

Helm value Default Description Environment variable

Helm value	Default	Description	Environment variable
kit: `operator.selfUpdate.enabled`; chart: `selfUpdate.enabled`	`true`	`false` leaves GKO’s image to Helm: GKO is listed in the UI, but its update is blocked. With `updatePolicy: gitops`, GKO never updates itself, whatever this value.	`GKO_SELF_UPDATE`
kit: `operator.selfUpdate.delaySeconds`; chart: `selfUpdate.delaySeconds`	`15`	Seconds without any update before GKO changes its own image.	`GKO_SELF_UPDATE_DELAY`
kit: `operator.selfUpdate.progressDeadlineSeconds`; chart: `selfUpdate.progressDeadlineSeconds`	`300`	`progressDeadlineSeconds` of the GKO Deployment. When it passes, the old GKO pod rolls the new one back.	none

kit: operator.selfUpdate.enabled; chart: selfUpdate.enabled

true

false leaves GKO’s image to Helm: GKO is listed in the UI, but its update is blocked. With updatePolicy: gitops, GKO never updates itself, whatever this value.

GKO_SELF_UPDATE

kit: operator.selfUpdate.delaySeconds; chart: selfUpdate.delaySeconds

15

Seconds without any update before GKO changes its own image.

GKO_SELF_UPDATE_DELAY

kit: operator.selfUpdate.progressDeadlineSeconds; chart: selfUpdate.progressDeadlineSeconds

300

progressDeadlineSeconds of the GKO Deployment. When it passes, the old GKO pod rolls the new one back.

none

The kit and the chart also set what the hand-over needs: the label com.molo17.conductor.type: operator, a RollingUpdate strategy with maxUnavailable: 0, minReadySeconds: 10, and leader election, with a Role rule on leases. GKO keeps the progress of its own update in the annotation gluesync.molo17.com/self-update of its Deployment.

Offline installs

With offline: true, GKO never calls the MOLO17 version catalog:

it shows no available version, and the UI cannot start a Core Hub update;
you set version and every module version in modules yourself;
the images must be in a registry the cluster can reach, usually the mirror set in imageRegistry.

An offline update still runs every step: maintenance mode, stability check and rollback.

kubectl -n gluesync-dev patch gsp gluesync --type merge \
  -p '{"spec":{"version":"2.3.2.2","modules":{"chronos":"0.3.3"}}}'

Disaster recovery

spec.disasterRecovery shapes what GKO builds when the Core Hub UI sets up a disaster-recovery pair (GKO 0.3.0 and later). Nothing in it turns disaster recovery on. Helm value: kit operator.disasterRecovery, chart platform.disasterRecovery.

Field Default Description

Field	Default	Description
`disasterRecovery.service.type`	`LoadBalancer`	Type of the Service `gko-disaster-recovery` that publishes ports 1718 and 1719: `LoadBalancer`, `NodePort`, or `ClusterIP` when you publish them yourself.
`disasterRecovery.service.annotations`, `loadBalancerClass`, `loadBalancerIP`, `loadBalancerSourceRanges`	empty	For the load balancer.
`disasterRecovery.service.sinkNodePort`, `controlNodePort`	allocated	Node ports of 1718 and 1719 for a `NodePort`.
`disasterRecovery.peerSink`, `peerControlURL`	empty	`GS_DR_PEER_SINK` and `GS_DR_PEER_CONTROL_URL`, when the other site publishes its ports elsewhere than 1718 and 1719 of the host of its URL.
`disasterRecovery.chronosDataOwner`	`true`	For a Core Hub with a `runAsUser`: `false` leaves out the root init container that gives the Chronos data to that user. Started as root, the Core Hub image does it itself.

disasterRecovery.service.type

LoadBalancer

Type of the Service gko-disaster-recovery that publishes ports 1718 and 1719: LoadBalancer, NodePort, or ClusterIP when you publish them yourself.

disasterRecovery.service.annotations, loadBalancerClass, loadBalancerIP, loadBalancerSourceRanges

empty

For the load balancer.

disasterRecovery.service.sinkNodePort, controlNodePort

allocated

Node ports of 1718 and 1719 for a NodePort.

disasterRecovery.peerSink, peerControlURL

empty

GS_DR_PEER_SINK and GS_DR_PEER_CONTROL_URL, when the other site publishes its ports elsewhere than 1718 and 1719 of the host of its URL.

disasterRecovery.chronosDataOwner

true

For a Core Hub with a runAsUser: false leaves out the root init container that gives the Chronos data to that user. Started as root, the Core Hub image does it itself.

See Disaster recovery on Kubernetes.

Status fields

GKO writes the status of the resource. The main fields:

Field Description

Field	Description
`phase`	`Current`, `UpdateAvailable`, `Updating`, `Degraded` or `Failed`. See Phases.
`version`, `available`	Core Hub version that runs, and the latest version on the release channel.
`releaseChannel`	Release channel read from Core Hub.
`components`	Each managed workload: name, kind, type, image, version, latest available version, and whether it is Ready.
`update`	The update in progress: target version, current step, the images recorded before the update, and the pipelines it put into maintenance mode. A restarted operator carries on from here.
`pausedPipelines`	Pipelines the running update has put into maintenance mode and not yet resumed.
`lastAttempt`	Outcome of the last update: result, message, and one result per component.
`history`	The last 10 updates, newest first.
`message`	Last progress or error message.
`disasterRecovery`	The disaster-recovery setup GKO made: whether it is enabled and set up by GKO, the node id, the peer URL, the failover mode, the address of the Service, and `lastChange`, the outcome of the last setup, join or disable (GKO 0.3.0 and later).

phase

Current, UpdateAvailable, Updating, Degraded or Failed. See Phases.

version, available

Core Hub version that runs, and the latest version on the release channel.

releaseChannel

Release channel read from Core Hub.

components

Each managed workload: name, kind, type, image, version, latest available version, and whether it is Ready.

update

The update in progress: target version, current step, the images recorded before the update, and the pipelines it put into maintenance mode. A restarted operator carries on from here.

pausedPipelines

Pipelines the running update has put into maintenance mode and not yet resumed.

lastAttempt

Outcome of the last update: result, message, and one result per component.

history

The last 10 updates, newest first.

message

Last progress or error message.

disasterRecovery

The disaster-recovery setup GKO made: whether it is enabled and set up by GKO, the node id, the peer URL, the failover mode, the address of the Service, and lastChange, the outcome of the last setup, join or disable (GKO 0.3.0 and later).