← all posts

Configuring and Monitoring 6 GHz Standard Power on a Catalyst 9800, Programmatically

September 22, 2026 · by Brent Leekley · ~8 min read

A two-pane terminal. On the left, the documented permission surface on a Catalyst 9800 for 6 GHz Standard Power: one boolean on the 6 GHz RF profile, with no per-AP power mode, no configurable AFC server and no API to trigger a query; separate writable RPCs handle AP geolocation. On the right, what you can read back: the show wireless afc ap readiness columns and two YANG operational models carrying grants, expiry, counters and health. Caption: thin permission, rich to monitor.

Last post in the 6 GHz power mode series. The first three covered the LPI versus Standard Power decision, the AFC machinery, and channel width. This one covers the operations column of that trade in the most practical way I can: what you can set, verify, and monitor on a Catalyst 9800 without touching a GUI.

Everything below is either from Cisco's published docs and YANG models or measured on my own bench, a 9800-CL running IOS-XE 17.17.1 under CML with NETCONF enabled. I will say which is which as we go, because the gap between what the documentation says and what the box actually does is most of the value here.

What can you actually configure?

The documented RF-profile permission for Standard Power is one boolean. Deployment involves more writable surface than that, but the permission itself really is this narrow.

Cisco shipped AFC support in IOS-XE 17.12.3. The RF-profile control for 6 GHz Standard Power is a single on/off bit: allow it, or do not. There is no per-AP power mode setting on that bit, no way to force a specific radio into Standard Power from it, no configurable AFC server, and no API that triggers an AFC query. You allow Standard Power on a profile, and the controller and the AFC service decide the rest.

That narrowness is less alarming than it sounds once you accept what AFC is: the database makes the channel and power decision, so there is not much for you to turn on that axis. But "one boolean" describes the permission, not the whole deployment. Geolocation height and its uncertainty, and AP location updates, are separate writable RPCs on the same controller, covered a few paragraphs below, and getting AFC to grant anything depends on those being filled in correctly. There is very little to automate on the permission itself and a great deal to automate on geolocation and monitoring, which is backwards from how most network automation projects start.

The permission bit also has a different name on every surface, and none of the three docs mention the other two names.

SurfaceName
C9800 CLItx-power standard, under ap dot11 6ghz rf-profile
IOS-XE YANGstd-pwr-mode-allowed, in Cisco-IOS-XE-wireless-rf-cfg (present from 17.13.1)
Catalyst Center Intent APIenableStandardPowerService, on the RF profile endpoints

It is the same part with a different part number in the catalog, the shop manual, and on the box, and nobody printed a cross-reference. I verified the first two on the bench. The on-box schema, pulled over NETCONF with get-schema, carries std-pwr-mode-allowed as a boolean defaulting to false, and show running-config all prints the default as no tx-power standard inside the stock 6 GHz RF profile. If you have ever grepped for one of these names, found nothing, and concluded the feature was missing, the likely cause is that you were searching one surface with another surface's name.

The rest of the writable surface is split across two modules, and the names on the RPC side are misleading enough to be worth walking through. They live in Cisco-IOS-XE-wireless-access-point-cmd-rpc, the same module that carries the experimental power-mode input from the end of this post. set-lrad-geolocation-height takes a mandatory height and height-uncertainty pair plus a mandatory choice of ap-name or mac-addr. ap-location-update is the one RPC in there that accepts coordinates, and its src leaf is constrained to the auto-location sources, so you cannot hand it a GNSS fix. The one that sounds like it writes a GNSS position, set-lrad-geolocation-gnss, does not: its only payload leaf is cable-len, the length of the GNSS external antenna cable.

There is also a plain configuration path, which matters more than the RPCs for most deployments. Since 17.15.1 you can set AP height under the priming profile, wireless profile ap priming name and then geolocation height H uncertainty U, modeled as the height, height-uncertainty and height-type leaves of ap-priming-profile in Cisco-IOS-XE-wireless-site-cfg. Config beats an RPC here for two practical reasons: it pre-provisions APs that have not joined yet, and it survives an AP flap.

There is a trap in that path worth knowing before you automate it. Cisco's AFC documentation states that the height and uncertainty are pushed to an AP only if priming-override is enabled, and the matching leaf in the same grouping, override-existing-priming, defaults to false. Leave it at the default and your config applies cleanly, the profile reads back exactly as you wrote it, and the AP height AFC needs is never actually set. Set it explicitly in whatever you generate. AFC still decides the frequencies and the power from whatever you send it, and nothing on the controller lets you skip that decision or watch it weigh its inputs.

One setting with three different names. On the Catalyst 9800 CLI it is tx-power standard under ap dot11 6ghz rf-profile. In IOS-XE YANG it is std-pwr-mode-allowed in Cisco-IOS-XE-wireless-rf-cfg, present from 17.13.1. In the Catalyst Center Intent API it is enableStandardPowerService on the RF profile endpoints. None of the three docs mentions the other two names.
Search one surface with another surface's name and the feature looks missing.

What can you monitor?

You can monitor considerably more, and this is where the useful work is.

The command show wireless afc ap has a column for every condition an AP must satisfy: AFC status, power mode capability, current power mode, AP and radio and RF-profile admin states, the tx-power standard bit, country allowed, location known, and height known. Read across a row and you have the readiness checklist for that AP, which is a much better troubleshooting experience than most features give you.

Over NETCONF and RESTCONF, two operational models carry the same story with more depth. Cisco-IOS-XE-wireless-afc-oper holds the AFC request and response per AP, including per-channel granted power and the grant expiry timestamp. That expiry timestamp is the one I would put on a dashboard first, because it is the clock from post two made visible.

Cisco-IOS-XE-wireless-afc-cloud-oper holds the service side: message counts, error counts, round-trip times, and a health check block that runs on a timer. My bench controller, still in its not-onboarded state, checks about every 30 seconds; Cisco's docs describe longer intervals for a healthy, onboarded controller, so treat the 30-second figure as what a failing check looks like here, not a general cadence. That same controller reports the precise blocker in the health block: not-otp-upgraded, true. Both models are operational data, so they are also candidates for streaming telemetry subscriptions. Whether they support on-change delivery or only periodic polling is the one question I have not been able to settle without a 6 GHz AP on the bench.

The readiness columns from show wireless afc ap. One row per access point covers AFC status, power mode capability, current power mode, the AP, radio and RF profile admin states, the tx-power standard bit, country allowed, location known and height known. Read across a row and you have the checklist for that AP.
One row per AP, and the two columns that usually fail indoors are location and height.

The gaps to plan around

Two findings from the research that produced this series are negative results, and they change tooling decisions, so they are worth stating as plainly as the features.

Catalyst Center exposes the Standard Power bit in its Intent API, but as of the 3.1.6 release, its native Intent API has no endpoint for provisioning the per-AP geolocation that AFC requires. I swept the full 3.1.6 endpoint documentation, 1,393 endpoints, and found no write path for geolocation or coordinates there, only a read field for AP height. That is a gap in Catalyst Center's own API, not a claim that geolocation is impossible to set by API at all: the controller-level NETCONF RPCs from the previous section are a separate, working API surface for exactly this, which is how my bench sets it. Cisco's newer 3.2.3 SDK also adds per-AP geolocation height and uncertainty to the Catalyst Center access point configuration write path, so check both your release and which API surface you are actually targeting before you build tooling around the gap.

genieparser, the parser library behind pyATS and Genie, has no parser for any show wireless afc command as of version 26.8. A CLI-scraping monitor means writing and maintaining your own parsers, which is a standing tax every time Cisco changes a column header. The YANG operational models above make that unnecessary, which is why the check below runs six of its seven sections over NETCONF, keeping one CLI section on purpose as the no-parser fallback pattern.

A working pyATS monitor

I built the check as a pyATS Blitz trigger: plain YAML, no Python, using the yang action against three models plus one execute action as a CLI cross-check. Five of the yang sections read the two AFC operational models, and the sixth is a get-config against Cisco-IOS-XE-wireless-rf-cfg, which is where the permission bit lives. That config-versus-operational split is the structural point of the check. It asserts that a get against the AFC model is accepted, that message and health-check error counters are zero, that the country is supported, that the AFC cloud health check reports healthy, and that the RF profile actually allows Standard Power.

One scope note before anyone builds on it. This is a controller-side precheck, and a green run means exactly those checks passed. It does not rule out every controller-side blocker, and it does not check per-AP location, capability, or grants, because those live in the show command and the oper model above. A few sharper limits worth knowing: the error counters are cumulative, so a controller that had a bad hour and recovered can still show nonzero counts, which is not the same as being unhealthy right now. The CLI section runs show wireless afc statistics and greps for two expected field labels, not for the readiness columns from earlier, and it passes on a controller with zero 6 GHz APs joined, so it is closer to a syntax check on that section than a fleet check. The first YANG get can return an empty payload without an error, which confirms the model is reachable but not that it is populated for your deployment. And the RF-profile name in the trigger is hard-coded, so it needs to match the actual profile at each site before that assertion means anything there.

Against my bench controller it reports five checks passing and two failing, and both failures are true statements: the controller is not OTP upgraded, and the Standard Power bit is still at its default of false. A precheck that passed everything on that box would be reporting a false result, which is a good reason to run a new monitor against a box you know is broken before you trust it against one you think is fine. Easypy gives you the HTML and JSON reporting without extra work, so the output drops into whatever review process you already run. I have not pasted a sanitized copy of that report here; the repo below is the place to see the real shape of a run, sanitize your own, and compare notes.

The testbed, trigger YAML, and job file are small, and I packaged them with the decision material from this series in a public repo: 6ghz-power-mode-toolkit. The precheck's own README covers these limits in more detail than fits here.

The pyATS Blitz precheck, seven sections. Five run over NETCONF against the two AFC operational models, a sixth is a get-config against the RF-profile config model, and one runs a CLI command on purpose as the no-parser fallback pattern. Against the bench controller five checks pass and two fail, and both failures are true: the controller is not OTP upgraded and the Standard Power bit is still at its default of false.
Six sections over NETCONF, one over CLI on purpose, run against a box known to be incomplete.

One loose end worth chasing

The on-box RPC model Cisco-IOS-XE-wireless-access-point-cmd-rpc, the same module with the geolocation RPCs from earlier, has carried an input leaf called rlp-sp-pwrmode-switch since 17.11.1, described in the schema as a recommendation to switch a radio from LPI to Standard Power. Cisco documents it nowhere that I can find. Straight from the published schema:

leaf rlp-sp-pwrmode-switch {
  type boolean;
  description
    "This is a recommendation to do power mode switch from LPI mode to SP mode";
}

The enclosing RPC is set-open-rrm-channel-auto, the Open RRM channel-assignment path, and on my 17.17.1 bench it is present in the on-box schema and its input is schema-validated: a misspelled name comes back as an unknown element, and the real name with missing leaves comes back as a missing element. I originally read that as proof of a live handler behind it. It is not. The NETCONF server validates against the loaded YANG modules before it dispatches to anything, so both of those responses are exactly what an advertised schema with no backend would produce. All the test really establishes is that the RPC is in the schema, which reading the module already told me. Proving there is a handler needs a fully populated, otherwise valid call and something on the controller that shows it arrived.

That validation is the part worth knowing before you go looking for a power-mode API. The rlp-sp-pwrmode-switch leaf is optional, but eighteen of its siblings in the same grouping are mandatory: slot ID, band, the primary and extension channels, both cost metrics, the two secondary channels, the three second-best fields, the channel widths, the redundancy and RRF fields, and the slave channel with its metric, plus a mandatory choice of ap-name or mac-addr. You cannot send the power-mode recommendation on its own. Any valid call is a complete channel assignment that happens to carry the recommendation along with it.

Here is a minimal NETCONF request for the geolocation-height RPC from earlier instead, built from the same schema, to show the shape of the surface:

<rpc xmlns="urn:ietf:params:xml:ns:netconf:base:1.0" message-id="101">
  <set-lrad-geolocation-height xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-wireless-access-point-cmd-rpc">
    <height>12</height>
    <height-uncertainty>3</height-uncertainty>
    <ap-name>AP-Floor3-12</ap-name>
  </set-lrad-geolocation-height>
</rpc>

What I could not test is the effect of the recommendation flag, because my joined APs have no 6 GHz radios. If you have a 9136 or 9166 on a lab controller and want to try it, go in knowing what you are actually sending. It is not a power-mode call, it is a channel assignment with a recommendation attached, and it will move the radio's channel as a side effect. Keep it on a lab box. It is still the most interesting unanswered question in this area, and I would like to compare notes.

The rest of the series

  1. Choosing a 6 GHz Power Mode: Low Power Indoor or Standard Power.
  2. What AFC Requires From a Standard Power Access Point.
  3. Why I Start at 40 MHz for Enterprise 6 GHz.
  4. Configuring and Monitoring 6 GHz Standard Power on a Catalyst 9800, Programmatically (this post).

← all posts