[DO NOT MERGE] feat: Update consumed cycles metrics on free cost schedule

[dsarlis]

Separate the values of NominalCycles that are used to count consumed cycles from the Cycles amounts that are charged against the canister's balance. Specifically, NominalCycles are updated even on a free schedule with the amounts that correspond to the respective resource usage.

The main change in production code is in the constructor of CompoundCycles while the rest are updating tests as needed, including fixing expected values, testing both on normal and free schedule as well as adding some new tests for cases that were not covered. Additionally, a const is defined for the canister creation fee to be consistently used in production and test code where necessary.

This should be deployed after #9676 to ensure that no wrong values are used before this PR separates the two types of cycles.

[Copilot]

Pull request overview

This PR updates cycles accounting so that consumed-cycles metrics (nominal) are updated based on resource usage even when the subnet runs on a free cost schedule, while charged cycles (real) remain zero where appropriate. It also adapts and expands tests to validate behavior on both normal and free schedules.

Changes:

• Update CompoundCycles::new to always preserve nominal cycles while applying cost-schedule rules only to the real (charged) portion.
• Adjust scheduler/execution/messaging tests to use .real() for balance-affecting paths and .nominal() for consumed-metrics assertions, including running key tests under both Normal and Free schedules.
• Centralize canister creation fee as a constant and update tests to assert nominal consumption on free schedules.

Reviewed changes

Copilot reviewed 11 out of 11 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
rs/types/cycles/src/compound_cycles.rs	Separates nominal vs real cycles in CompoundCycles::new, ensuring nominal updates even on free schedule.
rs/messaging/src/routing/stream_handler/tests.rs	Runs relevant stream handler tests under both cost schedules; updates lost-cycles expectations.
rs/execution_environment/src/scheduler/tests/metrics.rs	Updates consumed-cycles metric tests to assert nominal amounts; adds coverage for free schedule.
rs/execution_environment/src/scheduler/tests/ecdsa.rs	Adjusts ECDSA fee usage to pass real cycles where a payment is required.
rs/execution_environment/src/scheduler/tests/charging.rs	Updates resource charging assertions to use real cycles from CompoundCycles.
rs/execution_environment/src/scheduler/tests.rs	Updates ECDSA injection to use real fee.
rs/execution_environment/src/scheduler/test_utilities.rs	Changes fee/cost helpers to return CompoundCycles and adds helpers for time advancement and compute allocation cost.
rs/execution_environment/src/execution/response/tests.rs	Extends refund behavior test to both schedules and validates nominal consumed-cycles updates.
rs/execution_environment/src/execution_environment/tests.rs	Extends ingress charging test to validate consumed-cycles metrics alongside balance effects across schedules.
rs/execution_environment/src/canister_manager/tests.rs	Updates free-schedule canister creation tests to expect nominal consumption equal to creation fee.
rs/config/src/subnet_config.rs	Introduces CANISTER_CREATION_FEE constant and reuses it in cycles account manager config.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[maksymar]

hi @dsarlis ! please address minor comments from copilot, otherwise LGTM.

[github-actions]

Run on 4757f48 URL: https://github.com/dfinity/ic/actions/runs/24125024071

Review dismissed by automation script.

[dsarlis]

@cgundy or @maksymar please kick off CI again to ensure a green run so that the PR is ready to be merged when the time is right (I marked the title on purpose to avoid accidental merge before the change we need is scheduled for rollout).

[github-actions]

Run on b65b63f URL: https://github.com/dfinity/ic/actions/runs/24126190790

[cgundy]

Looks like there are still some linting errors:

cargo fmt -- --check
info: syncing channel updates for 1.93.1-x86_64-unknown-linux-gnu
info: latest update on 2026-02-12 for version 1.93.1 (01f6ddf75 2026-02-11)
info: downloading 7 components
Diff in /__w/ic/ic/rs/execution_environment/src/scheduler/tests/metrics.rs:1219:
             Some(CanisterStatusType::Stopped),
         );
 
-        let removed_cycles = CompoundCycles::<Instructions>::new(
-            Cycles::from(1000_u128),
-            cost_schedule,
-        );
+        let removed_cycles =
+            CompoundCycles::<Instructions>::new(Cycles::from(1000_u128), cost_schedule);
         test.canister_state_mut(canister_id)
             .system_state
             .consume_cycles(removed_cycles);

[dsarlis]

Looks like there are still some linting errors:

cargo fmt -- --check
info: syncing channel updates for 1.93.1-x86_64-unknown-linux-gnu
info: latest update on 2026-02-12 for version 1.93.1 (01f6ddf75 2026-02-11)
info: downloading 7 components
Diff in /__w/ic/ic/rs/execution_environment/src/scheduler/tests/metrics.rs:1219:
             Some(CanisterStatusType::Stopped),
         );
 
-        let removed_cycles = CompoundCycles::<Instructions>::new(
-            Cycles::from(1000_u128),
-            cost_schedule,
-        );
+        let removed_cycles =
+            CompoundCycles::<Instructions>::new(Cycles::from(1000_u128), cost_schedule);
         test.canister_state_mut(canister_id)
             .system_state
             .consume_cycles(removed_cycles);

Yeah that's because I committed the suggestion and it happened to create a lint issue this time. I fixed it along with an update to the test that was needed. Should be fine now.

[github-actions]

Run on 2fa9f6e URL: https://github.com/dfinity/ic/actions/runs/24129011066