Release 2.1.0
| Release Number | 2.1.0 |
|---|---|
| Release Type | Feature |
| Release Date | November 18th, 2026 |
| Tag | 2.1.0 |
After upgrading, you can load two more systems into Infrahub with a sync project: Webex locations, workspaces and devices, and Microsoft Entra ID users and groups. Relationships that hold several peers now match the source after a sync, so a peer removed in the source is also removed in Infrahub. A diff run between scheduled syncs no longer forces the next incremental sync to read every resource in full.
Sync now removes peers that were added outside sync from a mapped relationship that holds several peers.
- If you add peers by hand to a relationship that a sync project maps, review those relationships before the next sync.
See Upgrade notes for details.
Keep your Webex inventory in Infrahubβ
Compare the locations, workspaces and devices registered in Webex with your source of truth, and load them into Infrahub with their peer relationships. The webex adapter is a source only: every data request uses GET, and the adapter never changes data in Webex.
- Start from the inventory example, which maps
/v1/locations,/v1/workspacesand/v1/devicestoWebexLocation,WebexWorkspaceandWebexDevice. Each workspace references its location and each device references its workspace. - Use a fixed access token (
auth_method: token) for short runs, or an authorized Service App (auth_method: service_app) for scheduled runs. The adapter renews Service App tokens in memory for the current process. - Set credentials through environment variables only. The
token_env,client_id_env,client_secret_envandrefresh_token_envsettings take the names of the variables;{env: ...}references are accepted only inorg_idand in query parameter values. - Read any other Webex JSON API resource by adding it to the configuration. The adapter follows Webex pagination until no next page remains.
- Set
max_retry_delay(default 300 seconds) to cap the wait between retries. When Webex returns aRetry-Afterlonger thanmax_retry_delay, the run stops instead of waiting.
See the Webex adapter page for the required scopes and every setting. (#401)
Load directory users and groups into Infrahubβ
Read the users, groups and group membership of a Microsoft Entra ID tenant through Microsoft Graph, and load them as nodes of the kinds defined in your Infrahub schema. The msgraphsync adapter is a source only and sends GET requests only.
- Install the optional packages first:
pip install msgraph-sdk azure-identity. - Authenticate as a registered application with
auth: client_secretin CI pipelines and on servers. Set the secret inMSGRAPH_CLIENT_SECRET; the adapter refuses aclient_secretkey in the configuration file. - Use
auth: defaulton a laptop afteraz login, or on an Azure host with a managed identity. - Choose
membership: direct(default) ormembership: transitiveto decide which group members are read. - Start from the example project.
See the Microsoft Graph adapter page for the application permissions and settings. (#396)
Bug fixesβ
Remove stale peers from relationships that hold several peersβ
Change a relationship in the source from A, B to A, C and the next sync leaves A, C in Infrahub. Previously the result was A, B, C, and repeating the sync never removed B. Sync now loads the existing peers before it compares them with the source, and removing every peer in the source also clears the relationship in Infrahub. When a desired peer cannot be found, sync logs it, adds the peers it can resolve and keeps the existing peers of that relationship. (#168)
Preview with diff without losing incremental syncsβ
Run diff between two syncs and the next sync --no-full-extract still reads only the changed records. Previously the diff run became the baseline, and the next sync read every destination kind in full. Only applied runs are used as the baseline now, and the log names the reason for each per-resource full extract. (#403)
Load hierarchies of the same kind from Infrahubβ
Load an Infrahub kind whose nodes reference other nodes of the same kind, such as a parent location, without an ObjectNotFound error. Peers excluded by your filters stay out of the sync. (#384)
Upgrade notesβ
Peers added outside syncβ
If: you add peers by hand in Infrahub to a relationship that holds several peers and that a sync project maps.
Then: review those relationships before the next sync, and add the peers to the source if they must stay. Sync now removes peers that are not in the source.
Notes: when a desired peer cannot be found in the store, sync keeps the existing peers of that relationship.
Full changelogβ
Addedβ
- Added a Webex source adapter with configurable read-only resources, token renewal, pagination and an example mapping locations, workspaces and devices with peer relationships. Default retry delays are capped by the
max_retry_delaysetting, aRetry-Afterlonger thanmax_retry_delaystops the run, and{env: ...}references are accepted inorg_idandparamsonly; the*_envcredential settings take environment variable names. (#401) - Added the
msgraphsyncsource adapter, which reads users, groups and group membership from Microsoft Entra ID through Microsoft Graph and loads them into Infrahub. The adapter needs the optional packagesmsgraph-sdkandazure-identity:pip install msgraph-sdk azure-identity. (#396)
Changedβ
- Sync now removes peers added outside sync from a mapped relationship that holds several peers, except when a desired peer cannot be found in the store. (#168)
Fixedβ
- During an Infrahub update, a relationship that holds several peers is now fetched before its peers are compared, so stale peers are removed. (#168)
- Running
diffno longer forces a full extract on the nextsync --no-full-extract. Only applied runs are used as the incremental baseline, and the reason for each per-resource full extract is logged. (#403) - Loading the Infrahub side of a sync now resolves relationships to nodes of the same kind before they are loaded, without including nodes excluded by filters. (#384)
- Updating an Infrahub node no longer fetches the schema from Infrahub, and the release pull request exemption from the bump-label check now requires the title to name the same version as its
release/<version>branch. (#366)
Housekeepingβ
- Releases are now prepared as a reviewable pull request instead of being committed straight to
main. Dispatchingtrigger-push-stable.ymlbumps the version, assemblesCHANGELOG.mdfrom the news fragments underchangelog/and opens achore(release)pull request; merging it creates the tag and publishes a GitHub Release whose body is the changelog section contributors wrote. Release Drafter, which generated release notes from pull request titles, has been removed. (#218) - Standardized release bump labels and now reject pull requests with missing or conflicting release intent. (#348)
- The changelog check now counts only newly added news fragments, so editing an existing fragment no longer satisfies it. The check's
pull_requesttrigger is disabled, so it does not run on pull requests yet. (#233)