Moving to the new patch management
What changes when you patch, which settings stay and what to check after the switch.
The new patch management keeps your existing policies. You do not need to start again, but some parts of your workflow change. This page explains what changes compared with the previous patch management and what to check after the switch.
No more finalizing cycles
Previously, manual approval meant deciding on every update and then finalizing the cycle. Now the cycle starts automatically in its scheduled maintenance window. Only approved updates install. Undecided updates wait and do not hold up the others.
Before it starts, the cycle is Upcoming and you can still change its decisions. The finalize and reopen actions are gone. New pending updates are added automatically; octoja checks the selection again before starting. Later policy changes do not change a cycle that has already started.
Important: If an approved update must not install this time, defer it for this cycle or cancel the upcoming cycle. Leaving a cycle unfinalized no longer stops installation.
Approvals last beyond one cycle
Previously, you made decisions for each cycle. Now octoja remembers your approvals and rejections for later regular cycles of the same ring:
Use Undo deferral to remove a one-cycle skip. Decisions apply to the specific update.
Your own decision takes priority over automatic rules, promotion and global approvals. Approve globally covers all policies and rings and can skip an age delay or promotion. Blocks and exclusions still apply. Decisions from cancelled cycles, emergency rollouts and Install now runs apply only to that run.
Promotion checks each update separately
Promotion lets an update move to the next ring after being tested in the previous one.
Previously, octoja waited for the cycle to finish and checked device results by operating system. Now it checks how that particular update did in the ring immediately before it. The rest of the cycle does not need to finish first.
For example, a driver fails in the pilot ring but a security update installs successfully. The security update can still move on if it meets your requirements.
You can still set a minimum success rate, a minimum device count and a no-failures requirement. If there are not enough results yet, the update stays Deferred. octoja checks it again before the next start. To let it through anyway, approve it directly in the destination ring. This replaces the old Force promote action.
Caution: Only devices in the previous ring that actually need the update provide results. If no device there needs it, or fewer devices than your Minimum device count, the update can never meet your requirements. It then stays Deferred in every cycle until you approve it in the destination ring or globally.
Automatic approval now gives you two choices:
- Without promotion: Rules approve updates independently in every ring.
- With promotion: Rules approve the first ring. Later rings wait for results from the previous ring.
The new All updates rule saves you listing individual categories. You can add a waiting period too. If an update has no release date, that wait now starts when octoja first sees it. For undated updates already known to octoja, it starts at migration.
What happens to my existing settings?
What stays the same: Your policies, customer assignments, rings, schedules, time zones, approval rules and exclusions. Your existing restart postponement settings stay too. Existing global KB blocks remain in place.
What to check after the switch:
- Manual approval stays manual. It is now called Manual decisions. If promotion was on, it stays on but now works per update as described above.
- Promotion is switched off for automatic policies. This lets rules keep approving independently in every ring. Your old thresholds are saved. Only turn promotion back on if later rings should wait for test results.
- Cycles that have not started are checked again. Previously finalized cycles return to Upcoming too. Approvals from rules or promotion are checked again and may change. You do not need to finalize again.
- Manual decisions are kept. Exceptions are old entries without a recorded user or decision time, and Windows entries with incomplete update identifiers. Review the affected updates after the switch.
Update your agents too: the new patch execution needs agent 1.1.0 or newer. Older agents receive no patch work.
More control over rings and installations
- Rings have their own tabs and can be reordered. Use Move ring earlier or Move ring later, then save. Their schedules do not move with them. Check that the new preceding ring runs early enough for the next one.
- More flexible day offsets: Instead of just the second or third Tuesday, you can now add 0–31 days to the first to fourth Monday, Tuesday, and so on, of the month. For example, the first Friday plus seven days. Existing schedules do not shift.
- Install now: Install selected updates immediately on one or several devices, even without a patch policy. This does not change approvals for later cycles or count toward promotion.
- Emergency rollouts are still available. You now explicitly choose an immediate or scheduled start. See Create an emergency rollout.
- Cancel run: You can stop runs that have already started. No further installations begin. An installation already in progress finishes and its result is kept. An open restart prompt stays active too.
For setup, see Set up patch policies and cycles.
Blocked updates stay visible
Previously, globally blocked KB updates were left out of new cycles. Now you see them as Blocked. Approving an update does not remove its block.
Reject in a ring affects only that ring. The catalog's Global rejections tab contains blocks instead. These can now match an update ID, category or title text, and apply to all customers or selected customers. Policy exclusions also support title text, a reason and an expiry date.
The same applies to known problems: if Exclude updates with known problems is on in the policy, an affected update stays blocked even if you approve it globally.
With Override safeguards, you can deliberately bypass a block or known problem for one update in one upcoming cycle. The blocking rule stays in place. The override applies to that cycle only; confirm it again in later cycles and other rings. For known problems, you can instead use Override known problems on the update's page to override them in every upcoming cycle of every policy and ring. If the reasons for blocking change, you must confirm again too. This needs the Override patch safeguards permission. Administrator groups get it automatically; other groups need it explicitly.
To see known problems only as warnings in general, turn off Exclude updates with known problems in the policy. To keep the switch on but stop a known problem from holding an update back indefinitely, turn on Limit how long known problems hold updates back and set a Maximum hold in days after release.
See Exclude or block updates for the steps and permissions.
Clearer restart settings
Under Restart behavior, you now get a preview: What the user sees. New settings let you choose the response time for the final prompt and whether to Wait for Restart now.
- Waiting off: The final prompt counts down, then the device restarts.
- Waiting on: After the allowed postponements, the prompt waits for the signed-in user to click. It returns to the front regularly. If nobody is signed in, the device restarts automatically.
After migration, the response time is 5 minutes and waiting is off. Your existing postponement count and interval stay the same. Even with zero postponements, the final warning still appears.
Restart and install more updates (maximum) only limits how often octoja continues installing after restarting. A value of 0 does not prevent a necessary restart. Turn off Allow restarts if octoja must not restart the device at all.
Unlike before, waiting for a restart no longer keeps the device run open beyond the maintenance window. No new installation starts after the window; an already accepted installation can still finish. The restart prompt stays active, even after cancellation. If the restart happens after the window, further updates wait for a later cycle.
Clearer results and filters
Completed does not mean successfully installed. If nothing may install when the window opens, the cycle now ends as Completed with a reason. This also happens when every approved update is blocked. Previously, an unfinalized manual cycle became Missed.
New runs and reports make remaining actions clearer:
- Awaiting Restart on Windows: The final check after restarting is still outstanding.
- Awaiting User on macOS: The user must finish installing the update in Software Update.
- No data instead of 0% for Patch Compliance: The selection has no eligible Windows or macOS devices. Devices without inventory data do not count as compliant.
On a device, approval, blocks and the next installation are shown separately from the last attempt. You can view results by device or by update. See Review patch results and the report reference.
Devices, Cases and the Update Catalog have consistent filters. On Devices, you can now choose With patch policy or Without patch policy. This does not select a particular policy. In the catalog, you can filter by customer, platform, severity and global decision, among other options. Microsoft Defender's own security intelligence and platform updates are no longer listed, because Defender installs them itself.
The community blocking signal has changed too: only active KB blocks for all customers count. Other blocks and policy exclusions do not. Changes appear after the next synchronization. The signal is not a guarantee that an update is safe. See Understand smart patch advisories.
Check before your first maintenance window
- Update agents and open your migrated policies.
- Check ring order, schedules, promotion and restarts.
- Review approvals in the upcoming cycle. Defer updates or cancel the cycle if they must not run this time.
- After the window, check individual update results, not just Completed.
More details
What happens to older approvals and results?
What happens to older approvals and results?
Manual decisions are kept if the update can still be identified and the user and decision time were recorded. Approvals and rejections then apply to later cycles of the same ring too; deferrals still apply only to their original cycle.
Running and past cycles remain in the history. However, a cycle shown as running returns to Upcoming if no work had been sent to devices and there were no results.
Some old Windows approvals and results with incomplete update identifiers are removed. Old placeholder entries for updates that did not apply, or that rules declined, are removed too. Historical counts may therefore decrease. Update agents, refresh inventory and review newly reported updates.
Old Requires Reboot results become Succeeded after migration. This changes the old display; it does not mean a restart has just been verified.
Which results count toward promotion?
Which results count toward promotion?
Each update needs at least one successful installation in the immediately preceding ring and must meet your thresholds. For each device, octoja uses the latest matching result from that ring's regular cycles, including runs still in progress or cancelled after starting.
Succeeded and Awaiting Restart count as successful installation results. Failures count once the device has finished its run. Skipped, already installed, not applicable and unreported updates do not count toward the minimum device count. Emergency rollouts and Install now runs do not count either.
For the full explanation, see How patch management works.
Was this helpful?
More in Patch Management
How patch management worksSet up patch policies and cyclesDecide updates in patch cyclesReview patch resultsStill need help? Ask the team