Monitor mode

api/api_helpers.py

@@ -331,7 +331,7 @@ def get_phase_stats(ids):
     query = """
             SELECT
                 a.phase, a.metric, a.detail_name, a.value, a.type, a.max_value, a.min_value, a.unit,
-                b.uri, c.description, b.filename, b.commit_hash, b.branch
+                b.uri, c.description, b.filename, b.commit_hash, b.branch, b.monitor_run
             FROM phase_stats as a
             LEFT JOIN runs as b on b.id = a.run_id
             LEFT JOIN machines as c on c.id = b.machine_id
@@ -452,12 +452,14 @@ def get_phase_stats_object(phase_stats, case):
     for phase_stat in phase_stats:
         [
             phase, metric_name, detail_name, value, metric_type, max_value, min_value, unit,
-            repo, machine_description, filename, commit_hash, branch
+            repo, machine_description, filename, commit_hash, branch, monitor_run
         ] = phase_stat # unpack
 
         phase = phase.split('_', maxsplit=1)[1] # remove the 001_ prepended stuff again, which is only for ordering
 
-        if case == 'Repository':
+        if monitor_run:
+            key = 'monitor'
+        elif case == 'Repository':
             key = repo # Case D : RequirementsEngineering Case
         elif case == 'Branch':
             key = branch # Case C_3 : SoftwareDeveloper Case

docker/structure.sql

@@ -67,6 +67,7 @@ CREATE TABLE runs (
     logs text,
     invalid_run text,
     failed boolean DEFAULT false,
+    monitor_run boolean DEFAULT false,
     created_at timestamp with time zone DEFAULT now(),
     updated_at timestamp with time zone
 );

metric_providers/base.py

@@ -24,6 +24,8 @@ def __init__(
         sudo=False,
         disable_buffer=True,
         skip_check=False,
+        rootless=False,
+        monitor=False,
     ):
         self._metric_name = metric_name
         self._metrics = metrics
@@ -34,7 +36,8 @@ def __init__(
         self._sudo = sudo
         self._has_started = False
         self._disable_buffer = disable_buffer
-        self._rootless = None
+        self._monitor = monitor
+        self._rootless = rootless
         self._skip_check = skip_check
 
         self._tmp_folder = '/tmp/green-metrics-tool'
@@ -159,14 +162,18 @@ def start_profiling(self, containers=None):
             call_string += ' '  # space at start
             call_string += ' '.join(self._extra_switches)
 
+        if self._monitor is True:
+            call_string += ' --monitor '
         # This needs refactoring see https://github.com/green-coding-berlin/green-metrics-tool/issues/45
-        if (self._metrics.get('container_id') is not None) and (containers is not None):
+        elif (self._metrics.get('container_id') is not None) and (containers is not None):
             call_string += ' -s '
             call_string += ','.join(containers.keys())
 
         if self._rootless is True:
             call_string += ' --rootless '
 
+
+
         call_string += f" > {self._filename}"
 
         if platform.system() == "Linux":

metric_providers/cpu/time/cgroup/container/provider.py

@@ -3,13 +3,14 @@
 from metric_providers.base import BaseMetricProvider
 
 class CpuTimeCgroupContainerProvider(BaseMetricProvider):
-    def __init__(self, resolution, rootless=False, skip_check=False):
+    def __init__(self, resolution, rootless=False, skip_check=False, monitor=False):
         super().__init__(
             metric_name='cpu_time_cgroup_container',
             metrics={'time': int, 'value': int, 'container_id': str},
             resolution=resolution,
             unit='us',
             current_dir=os.path.dirname(os.path.abspath(__file__)),
             skip_check=skip_check,
+            rootless=rootless,
+            monitor=monitor,
         )
-        self._rootless = rootless

metric_providers/cpu/utilization/cgroup/container/provider.py

@@ -3,13 +3,14 @@
 from metric_providers.base import BaseMetricProvider
 
 class CpuUtilizationCgroupContainerProvider(BaseMetricProvider):
-    def __init__(self, resolution, rootless=False, skip_check=False):
+    def __init__(self, resolution, rootless=False, skip_check=False, monitor=False):
         super().__init__(
             metric_name='cpu_utilization_cgroup_container',
             metrics={'time': int, 'value': int, 'container_id': str},
             resolution=resolution,
             unit='Ratio',
             current_dir=os.path.dirname(os.path.abspath(__file__)),
-            skip_check = skip_check,
+            skip_check=skip_check,
+            rootless=rootless,
+            monitor=monitor,
         )
-        self._rootless = rootless

metric_providers/cpu/utilization/cgroup/container/source.c

@@ -6,10 +6,12 @@
 #include <time.h>
 #include <string.h> // for strtok
 #include <getopt.h>
+#include <dirent.h>
 
 typedef struct container_t { // struct is a specification and this static makes no sense here
     char path[BUFSIZ];
     char *id;
+    int active;
 } container_t;
 
 // All variables are made static, because we believe that this will
@@ -40,27 +42,57 @@ static long int read_cpu_cgroup(FILE *fd) {
     return cpu_usage;
 }
 
-static long int get_cpu_stat(char* filename, int mode) {
-    FILE* fd = NULL;
-    long int result=-1;
+static int scan_directory(container_t** containers, int rootless_mode) {
+    struct dirent* entry;
+    size_t docker_prefix_len = strlen("docker-");
+    size_t scope_suffix_len = strlen(".scope");
+    int length = 0;
+    DIR* dir = NULL;
+    char my_path[BUFSIZ] = "";
 
-    fd = fopen(filename, "r");
 
-    if ( fd == NULL) {
-        fprintf(stderr, "Error - Could not open path for reading: %s. Maybe the container is not running anymore? Are you using --rootless mode? Errno: %d\n", filename, errno);
-        exit(1);
-    }
-    if(mode == 1) {
-        result = read_cpu_cgroup(fd);
-        // printf("Got cgroup: %ld", result);
+    if(rootless_mode) {
+        sprintf(my_path, "/sys/fs/cgroup/user.slice/user-%d.slice/user@%d.service/user.slice/", user_id, user_id);
     } else {
-        result = read_cpu_proc(fd);
-        // printf("Got /proc/stat: %ld", result);
+        sprintf(my_path, "/sys/fs/cgroup/system.slice/");
     }
-    fclose(fd);
-    return result;
-}
 
+    dir = opendir(my_path);
+    if (!dir) {
+        fprintf(stderr,"Unable to scan directory for containers. Could not find folder: %s\n", my_path);
+        exit(-1);
+    }
+
+    *containers = malloc(sizeof(container_t));
+    //printf("old length: %d\n", length);
+
+    while ((entry = readdir(dir)) != NULL) {
+        // Check if the entry is a directory and matches the format
+        if (entry->d_type == DT_DIR &&
+            strstr(entry->d_name, "docker-") == entry->d_name &&
+            strstr(entry->d_name + docker_prefix_len, ".scope") != NULL &&
+            strcmp(entry->d_name + strlen(entry->d_name) - scope_suffix_len, ".scope") == 0) {
+            // printf("Entry %s\n", entry->d_name);
+            length++;
+            *containers = realloc(*containers, length * sizeof(container_t));
+            (*containers)[length-1].id = strdup(entry->d_name);
+            (*containers)[length-1].active = 1;
+            if(rootless_mode) {
+                sprintf((*containers)[length-1].path,
+                    "/sys/fs/cgroup/user.slice/user-%d.slice/user@%d.service/user.slice/%s/cpu.stat",
+                    user_id, user_id, entry->d_name);
+            } else {
+                sprintf((*containers)[length-1].path,
+                    "/sys/fs/cgroup/system.slice/%s/cpu.stat",
+                    entry->d_name);
+            }
+        }
+    }
+    //printf("Found new length: %d\n", length);
+
+    closedir(dir);
+    return length;
+}
 
 static int output_stats(container_t* containers, int length) {
 
@@ -69,28 +101,58 @@ static int output_stats(container_t* containers, int length) {
     long int cpu_readings_after[length];
     long int container_reading;
 
+    FILE *fd = NULL;
+
     struct timeval now;
     int i;
 
-
     // Get Energy Readings, set timestamp mark
     gettimeofday(&now, NULL);
+
     for(i=0; i<length; i++) {
-        //printf("Looking at %s ", containers[i].path);
-        cpu_readings_before[i]=get_cpu_stat(containers[i].path, 1);
+        // printf("Looking at %s ", containers[i].path);
+        fd = fopen(containers[i].path, "r");
+        if (fd == NULL) {
+            // printf("Warning, container has disappeared in 'before': %s\n", containers[i].path);
+            containers[i].active = 0;
+            continue;
+        }
+        containers[i].active = 1; // containers can come back up again and thus we set active = 1
+        cpu_readings_before[i]=read_cpu_cgroup(fd);
+        fclose(fd);
     }
-    main_cpu_reading_before = get_cpu_stat("/proc/stat", 0);
+
+    fd = fopen("/proc/stat", "r");
+    main_cpu_reading_before = read_cpu_proc(fd);
+    fclose(fd);
 
     usleep(msleep_time*1000);
 
+
+    // after the sleep the containers might have disappeared
+
     for(i=0; i<length; i++) {
-        cpu_readings_after[i]=get_cpu_stat(containers[i].path, 1);
+        if(containers[i].active == 0) continue;
+
+        fd = fopen(containers[i].path, "r");
+        if (fd == NULL) {
+            // printf("Warning, container has disappeared in 'after': %s\n", containers[i].path);
+            containers[i].active = 0;
+            continue;
+        }
+        cpu_readings_after[i]=read_cpu_cgroup(fd);
+        fclose(fd);
     }
-    main_cpu_reading_after = get_cpu_stat("/proc/stat", 0);
+
+    fd = fopen("/proc/stat", "r");
+    main_cpu_reading_after = read_cpu_proc(fd);
+    fclose(fd);
 
     // Display Energy Readings
     // This is in a seperate loop, so that all energy readings are done beforehand as close together as possible
     for(i=0; i<length; i++) {
+        if(containers[i].active == 0) continue;
+
         container_reading = cpu_readings_after[i] - cpu_readings_before[i];
         main_cpu_reading = main_cpu_reading_after - main_cpu_reading_before;
 
@@ -130,10 +192,11 @@ static int parse_containers(container_t** containers, char* containers_string, i
     int length = 0;
 
     for (; id != NULL; id = strtok(NULL, ",")) {
-        //printf("Token: %s\n", id);
+        // printf("Token: %s\n", id);
         length++;
         *containers = realloc(*containers, length * sizeof(container_t));
-        (*containers)[length-1].id = id;
+        (*containers)[length-1].id = strdup(id);
+        (*containers)[length-1].active = 1;
         if(rootless_mode) {
             sprintf((*containers)[length-1].path,
                 "/sys/fs/cgroup/user.slice/user-%d.slice/user@%d.service/user.slice/docker-%s.scope/cpu.stat",
@@ -163,7 +226,7 @@ static int check_system(int rootless_mode) {
         file_path_cpu_stat = "/sys/fs/cgroup/system.slice/cpu.stat";
     }
     file_path_proc_stat = "/proc/stat";
-    
+
     FILE* fd = NULL;
 
     fd = fopen(file_path_cpu_stat, "r");
@@ -192,6 +255,9 @@ static int check_system(int rootless_mode) {
 int main(int argc, char **argv) {
 
     int c;
+    int length = 0;
+    int discover_containers = 0;
+    int use_containers_string = 0;
     int check_system_flag = 0;
     int rootless_mode = 0; // docker root is default
     char *containers_string = NULL;  // Dynamic buffer to store optarg
@@ -207,11 +273,12 @@ int main(int argc, char **argv) {
         {"help", no_argument, NULL, 'h'},
         {"interval", no_argument, NULL, 'i'},
         {"containers", no_argument, NULL, 's'},
+        {"monitor", no_argument, NULL, 'm'},
         {"check", no_argument, NULL, 'c'},
         {NULL, 0, NULL, 0}
     };
 
-    while ((c = getopt_long(argc, argv, "ri:s:hc", long_options, NULL)) != -1) {
+    while ((c = getopt_long(argc, argv, "ri:s:hmc", long_options, NULL)) != -1) {
         switch (c) {
         case 'h':
             printf("Usage: %s [-i msleep_time] [-h]\n\n",argv[0]);
@@ -237,9 +304,13 @@ int main(int argc, char **argv) {
             rootless_mode = 1;
             break;
         case 's':
+            use_containers_string = 1;
             containers_string = (char *)malloc(strlen(optarg) + 1);  // Allocate memory
             strncpy(containers_string, optarg, strlen(optarg));
             break;
+        case 'm':
+            discover_containers = 1;
+            break;
         case 'c':
             check_system_flag = 1;
             break;
@@ -249,14 +320,21 @@ int main(int argc, char **argv) {
         }
     }
 
-    if(check_system_flag){
-        exit(check_system(rootless_mode)); 
+    if (discover_containers == 1 && use_containers_string == 1) {
+        fprintf(stderr,"Cannot discover containers and use containers string in parallel. Please choose either or.\n");
+        exit(-1);
+    } else if (use_containers_string == 1) {
+        length = parse_containers(&containers, containers_string, rootless_mode);
     }
 
-    int length = parse_containers(&containers, containers_string, rootless_mode);
+    if(check_system_flag){
+        exit(check_system(rootless_mode));
+    }
 
     while(1) {
+        if(discover_containers == 1) length = scan_directory(&containers, rootless_mode);
         output_stats(containers, length);
+        if(discover_containers == 1) free(containers);
     }
 
     free(containers); // since tools is only aborted by CTRL+C this is never called, but memory is freed on program end

metric_providers/memory/total/cgroup/container/provider.py

@@ -3,13 +3,14 @@
 from metric_providers.base import BaseMetricProvider
 
 class MemoryTotalCgroupContainerProvider(BaseMetricProvider):
-    def __init__(self, resolution, rootless=False, skip_check=False):
+    def __init__(self, resolution, rootless=False, skip_check=False, monitor=False):
         super().__init__(
             metric_name='memory_total_cgroup_container',
             metrics={'time': int, 'value': int, 'container_id': str},
             resolution=resolution,
             unit='Bytes',
             current_dir=os.path.dirname(os.path.abspath(__file__)),
             skip_check=skip_check,
+            rootless=rootless,
+            monitor=monitor,
         )
-        self._rootless = rootless

metric_providers/network/io/cgroup/container/provider.py

@@ -3,13 +3,14 @@
 from metric_providers.base import BaseMetricProvider
 
 class NetworkIoCgroupContainerProvider(BaseMetricProvider):
-    def __init__(self, resolution, rootless=False, skip_check=False):
+    def __init__(self, resolution, rootless=False, skip_check=False, monitor=False):
         super().__init__(
             metric_name='network_io_cgroup_container',
             metrics={'time': int, 'value': int, 'container_id': str},
             resolution=resolution,
             unit='Bytes',
             current_dir=os.path.dirname(os.path.abspath(__file__)),
             skip_check=skip_check,
+            rootless=rootless,
+            monitor=monitor,
         )
-        self._rootless = rootless

metric_providers/powermetrics/provider.py

@@ -41,7 +41,7 @@ def __init__(self, resolution, skip_check=False):
             ]
 
 
-    def check_system(self):
+    def check_system(self, check_command="default", check_error_message=None, check_parallel_provider=True):
         # no call to super().check_system() as we have different logic of finding the process
         if self._pm_process_count > 0:
             raise MetricProviderConfigurationError('Another instance of powermetrics is already running on the system!\n'

migrations/2024_05_15_monitor_mode.sql

@@ -0,0 +1 @@
+ALTER TABLE "runs" ADD COLUMN "monitor_run" boolean DEFAULT false;