commit bcad55d00c477a832cd244e1faf1dd15c1e0eeab
author David Tang <davtang@google.com> Fri Oct 31 15:17:53 2025 -0700
committer Copybara-Service <copybara-worker@google.com> Fri Oct 31 15:18:36 2025 -0700
tree 263fd1ceaf5e9358d8559104505f3415f0d7e4bf
parent 84bdc9acfe98233c6fdf1a64eaf5f14df823fffc

Skip configuring sensors that have failed creation.

The sensor collector will now also skip reconfiguration of sampling for sensors with status CREATION_PENDING or CREATION_FAILED. These statuses will be logged. In either case, this indicates that the sensor is present in the tlBMC store, but has a bad status. With the previous logic, this causes the HFT stream to immediately terminate since it will attempt to access the sensor in the `sensor_key_to_task_id` map, causing a segfault.

Instead of ending the HFT stream, this change will allow us to detect the sensor as STATUS_MISSING in the HFT stream and allow debugging from the status message. Other valid sensors in the subscription will stream their readings as expected.

#tlbmc-hft

PiperOrigin-RevId: 826647508
Change-Id: Idc91966165deb3f0bf2aee6f9564352a7632da3d

tlbmc/collector/sensor_collector.cc

diff --git a/tlbmc/collector/sensor_collector.cc b/tlbmc/collector/sensor_collector.cc
index 840ba8d..70b0ead 100644
--- a/tlbmc/collector/sensor_collector.cc
+++ b/tlbmc/collector/sensor_collector.cc
@@ -735,10 +735,13 @@
     // board config.
     if (config.key.empty()) {
       for (auto& [key, sensor] : key_to_sensor) {
-        if (sensor->GetSensorAttributesDynamic().state().status() ==
-            STATUS_CREATION_PENDING) {
+        if (Status sensor_status =
+                sensor->GetSensorAttributesDynamic().state().status();
+            sensor_status == STATUS_CREATION_PENDING ||
+            sensor_status == STATUS_CREATION_FAILED) {
           LOG(WARNING) << "Skipping configuration for sensor " << key
-                       << " because it is pending creation.";
+                       << " because sensor status is: "
+                       << Status_Name(sensor_status);
           continue;
         }
         absl::Duration interval = get_sampling_interval(sensor, config);
@@ -759,10 +762,13 @@
     // given key.
     auto it = key_to_sensor.find(config.key);
     if (it != key_to_sensor.end()) {
-      if (it->second->GetSensorAttributesDynamic().state().status() ==
-          STATUS_CREATION_PENDING) {
+      if (Status sensor_status =
+              it->second->GetSensorAttributesDynamic().state().status();
+          sensor_status == STATUS_CREATION_PENDING ||
+          sensor_status == STATUS_CREATION_FAILED) {
         LOG(WARNING) << "Skipping configuration for sensor " << config.key
-                     << " because it is pending creation.";
+                     << " because sensor status is: "
+                     << Status_Name(sensor_status);
         return absl::OkStatus();
       }
       absl::Duration interval = get_sampling_interval(it->second, config);