proc.plugin: add pressure stall information (#7209)

* proc.plugin: add pressure stall information

* dashboard_info: add "Pressure" section

* proc.plugin: mention PSI collector in doc

* dashboard_info: fix grammar in PSI section

* proc_pressure: fix wrong line name for "full" metrics

* proc_pressure: fix copypasta

* proc_pressure: refactor to prepare for cgroup changes

* cgroups.plugin: add pressure monitoring

* add proc_pressure.h to targets

* Makefile.am: fix indentation

* cgroups.plugin: remove a useless comment

* cgroups.plugin: fix pressure config name

* proc.plugin: arrange pressure charts under corresponding sections

* dashboard_info: rearrange pressure chart descriptions

* dashboard_info: reword PSI descriptions

5 files changed, 214 insertions, 0 deletions

diff --git a/collectors/proc.plugin/README.md b/collectors/proc.plugin/README.md
index 7e2aa10969..b81dd82f1b 100644
--- a/collectors/proc.plugin/README.md
+++ b/collectors/proc.plugin/README.md
@@ -18,6 +18,7 @@
 -   `/proc/interrupts` (total and per core hardware interrupts)
 -   `/proc/softirqs` (total and per core software interrupts)
 -   `/proc/loadavg` (system load and total processes running)
+-   `/proc/pressure/{cpu,memory,io}` (pressure stall information)
 -   `/proc/sys/kernel/random/entropy_avail` (random numbers pool availability - used in cryptography)
 -   `/sys/class/power_supply` (power supply properties)
 -   `ipc` (IPC semaphores and message queues)
diff --git a/collectors/proc.plugin/plugin_proc.c b/collectors/proc.plugin/plugin_proc.c
index fcb1babcfa..c9eef2c257 100644
--- a/collectors/proc.plugin/plugin_proc.c
+++ b/collectors/proc.plugin/plugin_proc.c
@@ -21,6 +21,9 @@ static struct proc_module {
         { .name = "/proc/loadavg", .dim = "loadavg", .func = do_proc_loadavg },
         { .name = "/proc/sys/kernel/random/entropy_avail", .dim = "entropy", .func = do_proc_sys_kernel_random_entropy_avail },
 
+        // pressure metrics
+        { .name = "/proc/pressure", .dim = "pressure", .func = do_proc_pressure },
+
         // CPU metrics
         { .name = "/proc/interrupts", .dim = "interrupts", .func = do_proc_interrupts },
         { .name = "/proc/softirqs", .dim = "softirqs", .func = do_proc_softirqs },
diff --git a/collectors/proc.plugin/plugin_proc.h b/collectors/proc.plugin/plugin_proc.h
index cb9a0c5f34..0c8dbaf47c 100644
--- a/collectors/proc.plugin/plugin_proc.h
+++ b/collectors/proc.plugin/plugin_proc.h
@@ -40,6 +40,7 @@ extern int do_proc_net_rpc_nfsd(int update_every, usec_t dt);
 extern int do_proc_sys_kernel_random_entropy_avail(int update_every, usec_t dt);
 extern int do_proc_interrupts(int update_every, usec_t dt);
 extern int do_proc_softirqs(int update_every, usec_t dt);
+extern int do_proc_pressure(int update_every, usec_t dt);
 extern int do_sys_kernel_mm_ksm(int update_every, usec_t dt);
 extern int do_sys_block_zram(int update_every, usec_t dt);
 extern int do_proc_loadavg(int update_every, usec_t dt);
@@ -66,6 +67,7 @@ extern void netdev_rename_device_add(const char *host_device, const char *contai
 extern void netdev_rename_device_del(const char *host_device);
 
 #include "proc_self_mountinfo.h"
+#include "proc_pressure.h"
 #include "zfs_common.h"
 
 #else // (TARGET_OS == OS_LINUX)
diff --git a/collectors/proc.plugin/proc_pressure.c b/collectors/proc.plugin/proc_pressure.c
new file mode 100644
index 0000000000..64974ae2eb
--- /dev/null
+++ b/collectors/proc.plugin/proc_pressure.c
@@ -0,0 +1,177 @@
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#include "plugin_proc.h"
+
+#define PLUGIN_PROC_MODULE_PRESSURE_NAME "/proc/pressure"
+#define CONFIG_SECTION_PLUGIN_PROC_PRESSURE "plugin:" PLUGIN_PROC_CONFIG_NAME ":" PLUGIN_PROC_MODULE_PRESSURE_NAME
+
+// linux calculates this every 2 seconds, see kernel/sched/psi.c PSI_FREQ
+#define MIN_PRESSURE_UPDATE_EVERY 2
+
+
+static struct pressure resources[PRESSURE_NUM_RESOURCES] = {
+        {
+            .some = { .id = "cpu_pressure", .title = "CPU Pressure" },
+        },
+        {
+            .some = { .id = "memory_some_pressure", .title = "Memory Pressure" },
+            .full = { .id = "memory_full_pressure", .title = "Memory Full Pressure" },
+        },
+        {
+            .some = { .id = "io_some_pressure", .title = "I/O Pressure" },
+            .full = { .id = "io_full_pressure", .title = "I/O Full Pressure" },
+        },
+};
+
+static struct {
+    procfile *pf;
+    const char *name;       // metric file name
+    const char *family;     // webui section name
+    int section_priority;
+} resource_info[PRESSURE_NUM_RESOURCES] = {
+        { .name = "cpu",    .family = "cpu",    .section_priority = NETDATA_CHART_PRIO_SYSTEM_CPU },
+        { .name = "memory", .family = "ram",    .section_priority = NETDATA_CHART_PRIO_SYSTEM_RAM },
+        { .name = "io",     .family = "disk",   .section_priority = NETDATA_CHART_PRIO_SYSTEM_IO },
+};
+
+void update_pressure_chart(struct pressure_chart *chart) {
+    rrddim_set_by_pointer(chart->st, chart->rd10, (collected_number)(chart->value10 * 100));
+    rrddim_set_by_pointer(chart->st, chart->rd60, (collected_number) (chart->value60 * 100));
+    rrddim_set_by_pointer(chart->st, chart->rd300, (collected_number) (chart->value300 * 100));
+
+    rrdset_done(chart->st);
+}
+
+int do_proc_pressure(int update_every, usec_t dt) {
+    int fail_count = 0;
+    int i;
+
+    static usec_t next_pressure_dt = 0;
+    static char *base_path = NULL;
+
+    update_every = (update_every < MIN_PRESSURE_UPDATE_EVERY) ? MIN_PRESSURE_UPDATE_EVERY : update_every;
+
+    if (next_pressure_dt <= dt) {
+        next_pressure_dt = update_every * USEC_PER_SEC;
+    } else {
+        next_pressure_dt -= dt;
+        return 0;
+    }
+
+    if (unlikely(!base_path)) {
+        base_path = config_get(CONFIG_SECTION_PLUGIN_PROC_PRESSURE, "base path of pressure metrics", "/proc/pressure");
+    }
+
+    for (i = 0; i < PRESSURE_NUM_RESOURCES; i++) {
+        procfile *ff = resource_info[i].pf;
+        int do_some = resources[i].some.enabled, do_full = resources[i].full.enabled;
+
+        if (unlikely(!ff)) {
+            char filename[FILENAME_MAX + 1];
+            char config_key[CONFIG_MAX_NAME + 1];
+
+            snprintfz(filename
+                      , FILENAME_MAX
+                      , "%s%s/%s"
+                      , netdata_configured_host_prefix
+                      , base_path
+                      , resource_info[i].name);
+
+            snprintfz(config_key, CONFIG_MAX_NAME, "enable %s some pressure", resource_info[i].name);
+            do_some = config_get_boolean(CONFIG_SECTION_PLUGIN_PROC_PRESSURE, config_key, CONFIG_BOOLEAN_YES);
+            resources[i].some.enabled = do_some;
+            if (resources[i].full.id) {
+                snprintfz(config_key, CONFIG_MAX_NAME, "enable %s full pressure", resource_info[i].name);
+                do_full = config_get_boolean(CONFIG_SECTION_PLUGIN_PROC_PRESSURE, config_key, CONFIG_BOOLEAN_YES);
+                resources[i].full.enabled = do_full;
+            }
+
+            ff = procfile_open(filename, " =", PROCFILE_FLAG_DEFAULT);
+            if (unlikely(!ff)) {
+                error("Cannot read pressure information from %s.", filename);
+                fail_count++;
+                continue;
+            }
+        }
+
+        ff = procfile_readall(ff);
+        resource_info[i].pf = ff;
+        if (unlikely(!ff)) {
+            continue;
+        }
+
+        size_t lines = procfile_lines(ff);
+        if (unlikely(lines < 1)) {
+            error("%s has no lines.", procfile_filename(ff));
+            fail_count++;
+            continue;
+        }
+
+        struct pressure_chart *chart;
+        if (do_some) {
+            chart = &resources[i].some;
+            if (unlikely(!chart->st)) {
+                chart->st = rrdset_create_localhost(
+                        "system"
+                        , chart->id
+                        , NULL
+                        , resource_info[i].family
+                        , NULL
+                        , chart->title
+                        , "percentage"
+                        , PLUGIN_PROC_NAME
+                        , PLUGIN_PROC_MODULE_PRESSURE_NAME
+                        , resource_info[i].section_priority + 40
+                        , update_every
+                        , RRDSET_TYPE_LINE
+                );
+                chart->rd10 = rrddim_add(chart->st, "some 10", NULL, 1, 100, RRD_ALGORITHM_ABSOLUTE);
+                chart->rd60 = rrddim_add(chart->st, "some 60", NULL, 1, 100, RRD_ALGORITHM_ABSOLUTE);
+                chart->rd300 = rrddim_add(chart->st, "some 300", NULL, 1, 100, RRD_ALGORITHM_ABSOLUTE);
+            } else {
+                rrdset_next(chart->st);
+            }
+
+            chart->value10 = strtod(procfile_lineword(ff, 0, 2), NULL);
+            chart->value60 = strtod(procfile_lineword(ff, 0, 4), NULL);
+            chart->value300 = strtod(procfile_lineword(ff, 0, 6), NULL);
+            update_pressure_chart(chart);
+        }
+
+        if (do_full && lines > 2) {
+            chart = &resources[i].full;
+            if (unlikely(!chart->st)) {
+                chart->st = rrdset_create_localhost(
+                        "system"
+                        , chart->id
+                        , NULL
+                        , resource_info[i].family
+                        , NULL
+                        , chart->title
+                        , "percentage"
+                        , PLUGIN_PROC_NAME
+                        , PLUGIN_PROC_MODULE_PRESSURE_NAME
+                        , resource_info[i].section_priority + 45
+                        , update_every
+                        , RRDSET_TYPE_LINE
+                );
+                chart->rd10 = rrddim_add(chart->st, "full 10", NULL, 1, 100, RRD_ALGORITHM_ABSOLUTE);
+                chart->rd60 = rrddim_add(chart->st, "full 60", NULL, 1, 100, RRD_ALGORITHM_ABSOLUTE);
+                chart->rd300 = rrddim_add(chart->st, "full 300", NULL, 1, 100, RRD_ALGORITHM_ABSOLUTE);
+            } else {
+                rrdset_next(chart->st);
+            }
+
+            chart->value10 = strtod(procfile_lineword(ff, 1, 2), NULL);
+            chart->value60 = strtod(procfile_lineword(ff, 1, 4), NULL);
+            chart->value300 = strtod(procfile_lineword(ff, 1, 6), NULL);
+            update_pressure_chart(chart);
+        }
+    }
+
+    if (PRESSURE_NUM_RESOURCES == fail_count) {
+        return 1;
+    }
+
+    return 0;
+}
diff --git a/collectors/proc.plugin/proc_pressure.h b/collectors/proc.plugin/proc_pressure.h
new file mode 100644
index 0000000000..3330218665
--- /dev/null
+++ b/collectors/proc.plugin/proc_pressure.h
@@ -0,0 +1,31 @@
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#ifndef NETDATA_PROC_PRESSURE_H
+#define NETDATA_PROC_PRESSURE_H
+
+#define PRESSURE_NUM_RESOURCES 3
+
+struct pressure {
+    int updated;
+    char *filename;
+
+    struct pressure_chart {
+        int enabled;
+
+        const char *id;
+        const char *title;
+
+        double value10;
+        double value60;
+        double value300;
+
+        RRDSET *st;
+        RRDDIM *rd10;
+        RRDDIM *rd60;
+        RRDDIM *rd300;
+    } some, full;
+};
+
+extern void update_pressure_chart(struct pressure_chart *chart);
+
+#endif //NETDATA_PROC_PRESSURE_H