In this Field Note
There is no prize for being the first WISP to discover why a firmware release needed another firmware release.
For an ordinary upgrade, I usually want a candidate to spend 30 to 90 days in the world before putting it across a working network. That is an operating preference, not a vendor recommendation or a security-patching deadline. The point is to collect evidence before asking paying customers to help with the test.
A security advisory can change that decision immediately. So can a reproducible fault that is already costing service. A useful firmware policy makes room for both patience and urgency and tells the person doing the work which one applies.
Give the upgrade a job
“A newer version exists” is a notification. It is not much of a change request.
Write one sentence explaining what the upgrade should accomplish. Perhaps it closes a vulnerability affecting your management interface, fixes a reproducible reboot, supports replacement hardware, or moves equipment onto a maintained branch. That sentence determines what you test and how long you can reasonably wait.
| Reason to change | A sensible next move |
|---|---|
| An affected, exposed service has an urgent security fix. | Follow the advisory, reduce exposure, and run an accelerated, recoverable deployment. |
| A known defect is hurting customers. | Reproduce it on a pilot and test whether the candidate actually resolves it. |
| A required feature or new device needs the release. | Test that dependency and the existing services it could disturb. |
| The release adds nothing you currently need. | Review support and security requirements; otherwise let useful operating experience accumulate. |
CISA’s Known Exploited Vulnerabilities Catalog is one input to the security decision. It identifies vulnerabilities with evidence of exploitation. Absence from that catalog does not establish that a release is safe to ignore; the vendor advisory and your exposure still matter.
A release has to earn its own reputation
The airMAX AC sequence from 8.7.22 through 8.7.23, 8.7.24, and 8.7.25 is the sort of run that gets my attention. I watched quick replacements and saw some intermediate downloads disappear. Following each candidate onto production equipment would have meant extra maintenance windows without much confidence to show for them.
That is my observation of the release cycle. It does not establish the same defect on every model, or explain every withdrawn download. Ubiquiti notes on its download site that security or regulatory requirements can also make releases unavailable.
When a candidate is replaced, start its observation clock again. A month of experience with one build does not age its replacement. Read reports for the exact hardware family and features you use. A complaint about a different platform may be irrelevant; three reports of the same required feature failing deserve investigation.
Keep the distinction between evidence and volume. A busy discussion thread is not a failure-rate study. Record the device, configuration, symptom, and whether anyone reproduced it. “People hate this release” is difficult to turn into a test.
Choose a pilot that can disappoint you usefully
A spare on the bench should establish that the image is correct, the configuration loads, management works, and recovery is possible. It cannot tell you much about a busy sector at 8 p.m.
The next pilot should represent the work: the relevant radio role, hardware revision, VLANs, bridging or routing, authentication, shaping, and management arrangement. It should also be reachable if the upgrade goes badly. Start with a device whose failure has a bounded consequence and a practical recovery route.
Ubiquiti’s firmware-management guide explains the platform distinctions and recovery references. Its advice is to run current stable firmware for security. My observation period is a risk-managed deployment practice, not a reason to disregard that advice indefinitely.
Before touching the pilot, save a short test card:
- Starting condition: approved version, configuration backup, uptime, link errors, radio retries, traffic, latency, and loss.
- Required behavior: the feature or defect being addressed, plus the services that must survive.
- Observation window: named busy periods or operating cycles, not merely an arbitrary number of hours.
- Stop conditions: lost management, unexplained reboots, failed subscriber service, or degradation outside the agreed baseline.
- Recovery: person responsible, approved image, access method, and latest point at which a site visit must begin.
A baseline should capture normal variation. Comparing a quiet Tuesday morning with a stormy Friday evening can produce a very persuasive explanation of the wrong problem.
Test the way back before you need it
A configuration export and a platform-specific backup may serve different recovery purposes. Keep the appropriate material off the device, protect credentials it contains, and establish which software and hardware can restore it. An unreadable backup with an encouraging filename remains an unreadable backup.
Check downgrade restrictions, configuration migrations, and bootloader requirements in the release notes. Do not promise rollback merely because an older image is available. If the change cannot be reversed cleanly, the plan needs a tested spare or another recovery method.
Order the work around management reachability. Changing both ends of your only path in one unattended operation can leave you with two devices to investigate and no route to either. Know which changes preserve access and who can reach the site if remote recovery ends.
Promote evidence, not enthusiasm
After the pilot, expand through small operating groups. Let each wave cover a meaningful load period, compare it with the baseline, and record exceptions. A deliberately approved mix of versions is easier to support than a fleet whose software depends on who last clicked the dashboard.
Set a review date for postponed upgrades. Otherwise “wait for evidence” quietly becomes “we have always used this version,” and the original judgment is lost.
If your first obstacle is an incomplete inventory or an untested recovery procedure, start there. Oaktown Tech can build a fleet-specific rollout and recovery plan with the people who will operate it. The handoff should make the next upgrade less adventurous and make “leave this candidate alone for another week” a defensible engineering decision.

Join the discussion