OpsChain 5.8.0: one approved change, every action under it
A change advisory board approves a change. It does not approve a run. Yet most automation records runs: the failed first attempt and its retry, four teams' workflows over a weekend, the reboot someone took by hand at 2 a.m. Each is a separate record, and the change the board approved has to be reconstructed from them afterwards, usually for an auditor, usually in a spreadsheet.
OpsChain 5.8.0 closes that gap. Everything in it serves one idea: the thing your board approved is the thing the platform holds, and everything done under it, by automation or by hand, lands on it.
1. Change records: one approved change, every run under it
opschain changes create opens a change record owned by your organisation or a project, with the scope of nodes the change may touch and, optionally, its window. A record moves through three states only: pending, approved, and closed with a verdict (completed, completed with incidents, rolled back, abandoned or cancelled).
While a record is approved, a run created inside its scope and its window joins it automatically. You can also name a record when you create or retry a run, and that always wins. Nested runs inherit their parent's record; triggered runs and retries inherit it while it is approved; a run can be linked after the fact, or unlinked with a reason when it joined the wrong change. Two approved records claiming the same node refuse the run by name until you say which one it belongs to, and a record still approved after its window has ended raises a warning rather than quietly staying open.
A record scoped to an environment now holds the workflow runs that change that environment, read from the node the workflow's steps actually target. A record that owns a maintenance window takes the work that window starts and admits.
opschain changes show renders the record's story: every run with its status and how it joined, every approval, attested step and break-glass act, in order, and filterable by kind so the record's own moments are not lost among the run events. opschain changes export writes the complete story as one file for your auditor, with principal ids only, and the server exports exactly the same bytes. Every run and every event names the change record it served.
2. The board approves the record, and production runs prove they belong
With Change Governance, an approval policy can hold a change record itself. Approving the record then requests the approval: the board sees the record in its inbox with its code, title, owner and scope, and the record is approved only when the policy's quorum is met, with the requester and the record's owner excluded from approving their own change.
A policy can also require that every run on its path belongs to an approved change record. A production run with none is refused at creation, by name, so change control is enforced at the door, not reconstructed afterwards. Run-level approval policies still gate every run; a record's approval never replaces them.
Two smaller changes make this easier to live with. The requester can withdraw a pending approval request with a reason; a withdrawal is not a decision, so a sole operator is never stuck between "self-approval is off" and "not an approver". And when the policy holding a record changes while a request is open, the request is replaced under the policy's current version, and the platform says so.
3. The human part of a change, attested
Not everything a change needs can be automated. A registrar, a vendor portal, a console setting, a change board's sign-off: these are acts a person performs. An action can now declare a manual step beside its automated ones: what the person does and where, the evidence it needs (an id the other system minted, a read-only check with its expected output, a file such as a screenshot, and the attester's statement), and who may attest it.
When the run reaches the step it holds. A permitted person attests it with the evidence and the run continues, or refuses it and the step fails by name. An unattested step past its timeout fails. Evidence files are sealed by digest beside the run, and a file whose content does not read as its declared type is refused. continue and blanket approvals never answer a manual step.
4. Break-glass records: the work done by hand, declared
Sometimes a run can only be rescued by hand, or a change has to be made outside any run. opschain breakglass record declares that act afterwards, on the run it rescued or the node it changed, with a mandatory reason, the act's window and the same four kinds of evidence. A record is correlated by its run, a maintenance window, an external change reference or a change record, and stays unreconciled until a person closes it with a note. An unreconciled record, and every file it names as evidence, is never pruned.
Break-glass records written on a workstation in local mode can now be exported and imported to the server, so your reviewers see them too.
5. Coverage: what claimed a node, and the ids the target minted
A step can record the identifiers the target system minted for what it changed: a cloud deployment id, a firewall rule id, a commit, an issue key. OpsChain never writes its own ids into the target system; it keeps the target's ids on the step.
Three reads use them. opschain governance coverage lists every change run, attested step and break-glass record that claimed a node and its subtree over a window, each marked as matchable exactly against the target's audit log by its id, or attributable by its window only. opschain governance activity exports the same activity across the estate. opschain governance references finds the step that reported an id you found in a cloud's audit log.
Every run also records the version of every setting it read, beside the commit it ran under, so a setting changed by hand between two runs is visible on the run that read the new value.
6. Your CAB's change in Jira Service Management drives the record
The Jira Service Management connector now meets change records. You declare once, on the connection, how your change statuses map onto pending, approved and closed, and which fields carry the class, window and scope. A mapping that leaves a status unmapped is refused when you save it, not discovered on the first change.
From then on the CAB drives the record. Creating the change opens a pending record; the CAB's approval approves it, and when an approval policy also holds the record, both must hold. Closing the change closes the record with the mapped verdict. A change the CAB pulls back holds the record so no new run joins it, while the runs already under it finish, and a close that arrives while work still runs waits for it.
In the other direction, every finished run, attested step, break-glass act and the close land on the change as internal notes, never customer-visible comments: run ids, outcomes and links, never a username, an email or a secret. If Jira is unreachable the notes are queued and retried; a run never waits for them.
7. Evidence that reaches your SIEM grouped by change
The SIEM forwarder exports the change record each event served on every lane: Splunk HEC, syslog/CEF and OpenTelemetry traces. Your security team can group a change's events in the tooling it already watches.
Also in this release
- Windows relative to now. Change records, break-glass records and maintenance windows accept windows such as
now/+2hor-30m/now, resolved on your clock, sent as UTC and printed back. - The runner protocol on the fleet view. The runner list names each runner's protocol, so a runner left on an older release after an upgrade is visible before work reaches it.
- A runner busy on a long step stays online in the runner list and placement explanations.
- A workflow's log includes the workflows it runs, nested at any depth, and withholds by name the nested runs a reader may not read.
- Files reach every step of a composed action, and governed image builds authenticate with an Azure certificate.
Getting it
OpsChain 5.8.0 is published with the same signed, verifiable distribution as every release: Developer ID and notarised on macOS, Authenticode on Windows, and a cosign-sealed checksum file under a pinned release key. The Install page carries every digest and fingerprint, none typed by hand.
Access to OpsChain is by invitation. Request access, or book a demo and ask to see a change record from the board's approval to the auditor's export.
Key takeaway
In 5.8.0 the change your board approved is the record the platform holds: every run, retry, attested manual step and break-glass act lands on it, it reads as one story and exports as one file, and your CAB's change in Jira Service Management opens, approves and closes it.
Ready to see OpsChain in action?
Book a personalised demo and see how OpsChain can transform your operations.
Book a DemoFounder & CEO, LimePoint
Goran is the founder of LimePoint and the creator of OpsChain. He is passionate about helping enterprises automate and govern their operations at scale.