Updates and fixes to recording rules subcommand of ceems_tool

.github/workflows/step_images.yml

@@ -87,7 +87,7 @@ jobs:
 
       - name: Push README to registry
         uses: christian-korneck/update-container-description-action@d36005551adeaba9698d8d67a296bd16fa91f8e8 # v1
-        if: (github.ref == 'refs/heads/main' || (github.event_name == 'push' && contains(github.ref, 'refs/tags/'))) && github.repository_owner == 'ceems' # Don't run this workflow on forks.
+        if: (github.ref == 'refs/heads/main' || (github.event_name == 'push' && contains(github.ref, 'refs/tags/'))) && github.repository_owner == 'ceems-dev' # Don't run this workflow on forks.
         env:
           # For dockerhub registry
           DOCKER_USER: ${{ secrets.login }}

CHANGELOG.md

@@ -1,11 +1,12 @@
 # Changelog
 
-## 0.11.0 / 2025-*-*
+## 0.11.0 / 2025-*
 
 ### Breaking Changes
 
 #### CEEMS Exporter
 
+- Collector `rapl` is disabled by default now and to enable it add `--collector.rapl` to CLI arguments.
 - Collector `ipmi_dcmi` has been renamed to `ipmi` as more functionality beyond DCMI has been added to the collector.
 - Following metric labels have been renamed to be more consistent with Prometheus naming convention:
     * `ceems_ipmi_dcmi_current_watts` -> `ceems_ipmi_dcmi_power_current_watts`
@@ -17,6 +18,18 @@
     * `ceems_redfish_max_watts` -> `ceems_redfish_power_max_watts`
     * `ceems_redfish_avg_watts` -> `ceems_redfish_power_avg_watts`
 
+#### CEEMS tool
+
+- Several minor bugs in recording rules have been fixed. Please regenerate the recording rules with new version of `ceems_tool`.
+- GPU profiling metrics have been renamed to have `prof` in the metric label. For instance, `uuid:ceems_gpu_sm_active:ratio` became
+`uuid:ceems_gpu_prof_sm_active:ratio`.
+- NVIDIA profiling metrics suffix has been corrected to use `sum` instead of `ratio` for NVLink, PCIe traffic metrics. Thus, metrics
+have been renamed as follows:
+    * `uuid:ceems_gpu_pcie_tx_bytes:ratio` -> `uuid:ceems_gpu_prof_pcie_tx_bytes:sum`
+    * `uuid:ceems_gpu_pcie_rx_bytes:ratio` -> `uuid:ceems_gpu_prof_pcie_rx_bytes:sum`
+    * `uuid:ceems_gpu_nvlink_tx_bytes:ratio` -> `uuid:ceems_gpu_prof_nvlink_tx_bytes:sum`
+    * `uuid:ceems_gpu_nvlink_rx_bytes:ratio` -> `uuid:ceems_gpu_prof_nvlink_rx_bytes:sum`
+
 ## 0.10.2 / 2025-08-07
 
 - [BUGFIX] Fix bpf code to work with LLVM 20 [#393](https://github.com/mahendrapaipuri/ceems/pull/393) ([@mahendrapaipuri](https://github.com/mahendrapaipuri))

cmd/ceems_tool/relabel.go

@@ -16,7 +16,7 @@ import (
 var gpuSeries = []string{
 	"DCGM_FI_DEV_POWER_USAGE_INSTANT",
 	"amd_gpu_power",
-	"gpu_power_usage",
+	"GPU_POWER_USAGE",
 }
 
 // MetricRelabelConfig contains the Prometheus metric relabel config.