Framework: Telemetry retake

Signed-off-by: pmikus <pmikus@cisco.com>
Change-Id: I2f019a083916aec9f7816266f6ad5b92dcc31fa0

4 files changed, 963 insertions, 0 deletions

diff --git a/resources/templates/telemetry/bpf_runtime.yaml b/resources/templates/telemetry/bpf_runtime.yaml
new file mode 100644
index 0000000000..88ad7eb64f
--- /dev/null
+++ b/resources/templates/telemetry/bpf_runtime.yaml
@@ -0,0 +1,130 @@
+---
+logging:
+  version: 1
+  formatters:
+    console:
+      format: '%(asctime)s - %(name)s - %(message)s'
+    prom:
+      format: '%(message)s'
+  handlers:
+    console:
+      class: logging.StreamHandler
+      level: INFO
+      formatter: console
+      stream: ext://sys.stdout
+    prom:
+      class: logging.handlers.RotatingFileHandler
+      level: INFO
+      formatter: prom
+      filename: /tmp/metric.prom
+      mode: w
+  loggers:
+    prom:
+      handlers: [prom]
+      level: INFO
+      propagate: False
+  root:
+    level: INFO
+    handlers: [console]
+scheduler:
+  duration: 1
+programs:
+  - name: bundle_bpf
+    metrics:
+      counter:
+        - name: cpu_cycle
+          documentation: Cycles processed by CPUs
+          namespace: bpf
+          labelnames:
+            - name
+            - cpu
+            - pid
+        - name: cpu_instruction
+          documentation: Instructions retired by CPUs
+          namespace: bpf
+          labelnames:
+            - name
+            - cpu
+            - pid
+        - name: llc_reference
+          documentation: Last level cache operations by type
+          namespace: bpf
+          labelnames:
+            - name
+            - cpu
+            - pid
+        - name: llc_miss
+          documentation: Last level cache operations by type
+          namespace: bpf
+          labelnames:
+            - name
+            - cpu
+            - pid
+    events:
+      - type: 0x0 # HARDWARE
+        name: 0x0 # PERF_COUNT_HW_CPU_CYCLES
+        target: on_cpu_cycle
+        table: cpu_cycle
+      - type: 0x0 # HARDWARE
+        name: 0x1 # PERF_COUNT_HW_INSTRUCTIONS
+        target: on_cpu_instruction
+        table: cpu_instruction
+      - type: 0x0 # HARDWARE
+        name: 0x2 # PERF_COUNT_HW_CACHE_REFERENCES
+        target: on_cache_reference
+        table: llc_reference
+      - type: 0x0 # HARDWARE
+        name: 0x3 # PERF_COUNT_HW_CACHE_MISSES
+        target: on_cache_miss
+        table: llc_miss
+    code: |
+      #include <linux/ptrace.h>
+      #include <uapi/linux/bpf_perf_event.h>
+
+      const int max_cpus = 256;
+
+      struct key_t {
+          int cpu;
+          int pid;
+          char name[TASK_COMM_LEN];
+      };
+
+      BPF_HASH(llc_miss, struct key_t);
+      BPF_HASH(llc_reference, struct key_t);
+      BPF_HASH(cpu_instruction, struct key_t);
+      BPF_HASH(cpu_cycle, struct key_t);
+
+      static inline __attribute__((always_inline)) void get_key(struct key_t* key) {
+          key->cpu = bpf_get_smp_processor_id();
+          key->pid = bpf_get_current_pid_tgid();
+          bpf_get_current_comm(&(key->name), sizeof(key->name));
+      }
+
+      int on_cpu_cycle(struct bpf_perf_event_data *ctx) {
+          struct key_t key = {};
+          get_key(&key);
+
+          cpu_cycle.increment(key, ctx->sample_period);
+          return 0;
+      }
+      int on_cpu_instruction(struct bpf_perf_event_data *ctx) {
+          struct key_t key = {};
+          get_key(&key);
+
+          cpu_instruction.increment(key, ctx->sample_period);
+          return 0;
+      }
+      int on_cache_reference(struct bpf_perf_event_data *ctx) {
+          struct key_t key = {};
+          get_key(&key);
+
+          llc_reference.increment(key, ctx->sample_period);
+          return 0;
+      }
+      int on_cache_miss(struct bpf_perf_event_data *ctx) {
+          struct key_t key = {};
+          get_key(&key);
+
+          llc_miss.increment(key, ctx->sample_period);
+          return 0;
+      }
diff --git a/resources/templates/telemetry/vpp_clear_stats.yaml b/resources/templates/telemetry/vpp_clear_stats.yaml
new file mode 100644
index 0000000000..9391502654
--- /dev/null
+++ b/resources/templates/telemetry/vpp_clear_stats.yaml
@@ -0,0 +1,171 @@
+---
+logging:
+  version: 1
+  formatters:
+    console:
+      format: '%(asctime)s - %(name)s - %(message)s'
+    prom:
+      format: '%(message)s'
+  handlers:
+    console:
+      class: logging.StreamHandler
+      level: INFO
+      formatter: console
+      stream: ext://sys.stdout
+    prom:
+      class: logging.handlers.RotatingFileHandler
+      level: INFO
+      formatter: prom
+      filename: /tmp/metric.prom
+      mode: w
+  loggers:
+    prom:
+      handlers: [prom]
+      level: INFO
+      propagate: False
+  root:
+    level: INFO
+    handlers: [console]
+scheduler:
+  duration: 1
+programs:
+  - name: bundle_vpp
+    metrics:
+      gauge:
+        - name: calls
+          documentation: Number of calls total
+          labelnames:
+            - name
+            - state
+            - thread_name
+            - thread_id
+            - thread_lcore
+        - name: vectors
+          documentation: Number of vectors total
+          labelnames:
+            - name
+            - state
+            - thread_name
+            - thread_id
+            - thread_lcore
+        - name: suspends
+          documentation: Number of suspends total
+          labelnames:
+            - name
+            - state
+            - thread_name
+            - thread_id
+            - thread_lcore
+        - name: clocks
+          documentation: Number of clocks total
+          labelnames:
+            - name
+            - state
+            - thread_name
+            - thread_id
+            - thread_lcore
+        - name: vectors_calls
+          documentation: Number of vectors per call
+          labelnames:
+            - name
+            - state
+            - thread_name
+            - thread_id
+            - thread_lcore
+    code: |
+      clear runtime
+  - name: bundle_vpp
+    metrics:
+      gauge:
+        - name: rx_packets
+          documentation: Number of received packets for interface
+          labelnames:
+            - name
+            - index
+        - name: rx_bytes
+          documentation: Number of received bytes for interface
+          labelnames:
+            - name
+            - index
+        - name: rx_error
+          documentation: Number of errors on interface
+          labelnames:
+            - name
+            - index
+        - name: tx_packets
+          documentation: Number of transitted packets for interface
+          labelnames:
+            - name
+            - index
+        - name: tx_bytes
+          documentation: Number of transitted bytes for interface
+          labelnames:
+            - name
+            - index
+        - name: tx_error
+          documentation: Number of errors on interface
+          labelnames:
+            - name
+            - index
+        - name: drops
+          documentation: Number of dropped packets for interface
+          labelnames:
+            - name
+            - index
+        - name: punt
+          documentation: Number of punted packets for interface
+          labelnames:
+            - name
+            - index
+        - name: ip4
+          documentation: Number of IPv4 packets for interface
+          labelnames:
+            - name
+            - index
+        - name: ip6
+          documentation: Number of IPv6 packets for interface
+          labelnames:
+            - name
+            - index
+        - name: rx_no_buf
+          documentation: Number of out of buffer RX packets on interface
+          labelnames:
+            - name
+            - index
+        - name: rx_miss
+          documentation: Number of missed RX packets on interface
+          labelnames:
+            - name
+            - index
+    code: |
+      clear interfaces
+  - name: bundle_vpp
+    metrics:
+      gauge:
+        - name: node_counter
+          documentation: Node counter
+          labelnames:
+            - name
+            - reason
+            - severity
+            - thread_name
+            - thread_id
+    code: |
+      clear node counters
+  - name: bundle_vpp
+    metrics:
+      gauge:
+        - name: calls
+          documentation: Load operations
+          labelnames:
+            - name
+            - thread_name
+            - thread_id
+        - name: packets
+          documentation: Load operations
+          labelnames:
+            - name
+            - thread_name
+            - thread_id
+    code: |
+      perfmon reset
diff --git a/resources/templates/telemetry/vpp_runtime.yaml b/resources/templates/telemetry/vpp_runtime.yaml
new file mode 100644
index 0000000000..4f0f6d7a7d
--- /dev/null
+++ b/resources/templates/telemetry/vpp_runtime.yaml
@@ -0,0 +1,508 @@
+---
+logging:
+  version: 1
+  formatters:
+    console:
+      format: '%(asctime)s - %(name)s - %(message)s'
+    prom:
+      format: '%(message)s'
+  handlers:
+    console:
+      class: logging.StreamHandler
+      level: INFO
+      formatter: console
+      stream: ext://sys.stdout
+    prom:
+      class: logging.handlers.RotatingFileHandler
+      level: INFO
+      formatter: prom
+      filename: /tmp/metric.prom
+      mode: w
+  loggers:
+    prom:
+      handlers: [prom]
+      level: INFO
+      propagate: False
+  root:
+    level: INFO
+    handlers: [console]
+scheduler:
+  duration: 1
+programs:
+  - name: bundle_vpp
+    metrics:
+      info:
+        - name: version
+          documentation: VPP version
+          namespace: vpp
+          subsystem: version
+          labelnames:
+            - version
+    code: |
+      show version
+  - name: bundle_vpp
+    metrics:
+      gauge:
+        - name: calls
+          documentation: Number of calls total
+          namespace: vpp
+          subsystem: runtime
+          labelnames:
+            - name
+            - state
+            - thread_name
+            - thread_id
+            - thread_lcore
+        - name: vectors
+          documentation: Number of vectors total
+          namespace: vpp
+          subsystem: runtime
+          labelnames:
+            - name
+            - state
+            - thread_name
+            - thread_id
+            - thread_lcore
+        - name: suspends
+          documentation: Number of suspends total
+          namespace: vpp
+          subsystem: runtime
+          labelnames:
+            - name
+            - state
+            - thread_name
+            - thread_id
+            - thread_lcore
+        - name: clocks
+          documentation: Number of clocks total
+          namespace: vpp
+          subsystem: runtime
+          labelnames:
+            - name
+            - state
+            - thread_name
+            - thread_id
+            - thread_lcore
+        - name: vectors_calls
+          documentation: Number of vectors per call
+          namespace: vpp
+          subsystem: runtime
+          labelnames:
+            - name
+            - state
+            - thread_name
+            - thread_id
+            - thread_lcore
+    code: |
+      clear runtime
+      wait {duration}
+      show runtime
+  - name: bundle_vpp
+    metrics:
+      gauge:
+        - name: rx_packets
+          documentation: Number of received packets for interface
+          namespace: vpp
+          subsystem: interface
+          labelnames:
+            - name
+            - index
+        - name: rx_bytes
+          documentation: Number of received bytes for interface
+          namespace: vpp
+          subsystem: interface
+          labelnames:
+            - name
+            - index
+        - name: rx_error
+          documentation: Number of errors on interface
+          namespace: vpp
+          subsystem: interface
+          labelnames:
+            - name
+            - index
+        - name: tx_packets
+          documentation: Number of transitted packets for interface
+          namespace: vpp
+          subsystem: interface
+          labelnames:
+            - name
+            - index
+        - name: tx_bytes
+          documentation: Number of transitted bytes for interface
+          namespace: vpp
+          subsystem: interface
+          labelnames:
+            - name
+            - index
+        - name: tx_error
+          documentation: Number of errors on interface
+          namespace: vpp
+          subsystem: interface
+          labelnames:
+            - name
+            - index
+        - name: drops
+          documentation: Number of dropped packets for interface
+          namespace: vpp
+          subsystem: interface
+          labelnames:
+            - name
+            - index
+        - name: punt
+          documentation: Number of punted packets for interface
+          namespace: vpp
+          subsystem: interface
+          labelnames:
+            - name
+            - index
+        - name: ip4
+          documentation: Number of IPv4 packets for interface
+          namespace: vpp
+          subsystem: interface
+          labelnames:
+            - name
+            - index
+        - name: ip6
+          documentation: Number of IPv6 packets for interface
+          namespace: vpp
+          subsystem: interface
+          labelnames:
+            - name
+            - index
+        - name: rx_no_buf
+          documentation: Number of out of buffer RX packets on interface
+          namespace: vpp
+          subsystem: interface
+          labelnames:
+            - name
+            - index
+        - name: rx_miss
+          documentation: Number of missed RX packets on interface
+          namespace: vpp
+          subsystem: interface
+          labelnames:
+            - name
+            - index
+    code: |
+      clear interfaces
+      wait {duration}
+      show interface
+  - name: bundle_vpp
+    metrics:
+      gauge:
+        - name: node_counter
+          documentation: Node counter
+          namespace: vpp
+          subsystem: counters
+          labelnames:
+            - name
+            - reason
+            - severity
+            - thread_name
+            - thread_id
+    code: |
+      clear node counters
+      wait {duration}
+      show node counters verbose
+  - name: bundle_vpp
+    metrics:
+      gauge:
+        - name: context_switches
+          documentation: Per-thread context switches
+          namespace: vpp
+          subsystem: context_switches
+          labelnames:
+            - name
+            - id
+    code: |
+      perfmon reset
+      perfmon start bundle context-switches
+      wait {duration}
+      perfmon stop
+      show perfmon statistics
+  - name: bundle_vpp
+    metrics:
+      gauge:
+        - name: minor_page_faults
+          documentation: Per-thread page faults (minor)
+          namespace: vpp
+          subsystem: page_faults
+          labelnames:
+            - name
+            - id
+        - name: major_page_faults
+          documentation: Per-thread page faults (major)
+          namespace: vpp
+          subsystem: page_faults
+          labelnames:
+            - name
+            - id
+    code: |
+      perfmon reset
+      perfmon start bundle page-faults
+      wait {duration}
+      perfmon stop
+      show perfmon statistics
+  - name: bundle_vpp
+    metrics:
+      gauge:
+        - name: calls
+          documentation: Instructions/packet, cycles/packet and IPC (calls)
+          namespace: vpp
+          subsystem: inst_and_clock
+          labelnames:
+            - name
+            - thread_name
+            - thread_id
+        - name: packets
+          documentation: Instructions/packet, cycles/packet and IPC (packets)
+          namespace: vpp
+          subsystem: inst_and_clock
+          labelnames:
+            - name
+            - thread_name
+            - thread_id
+        - name: packets_per_call
+          documentation: Instructions/packet, cycles/packet and IPC (packets/call)
+          namespace: vpp
+          subsystem: inst_and_clock
+          labelnames:
+            - name
+            - thread_name
+            - thread_id
+        - name: clocks_per_packets
+          documentation: Instructions/packet, cycles/packet and IPC (clocks/packets)
+          namespace: vpp
+          subsystem: inst_and_clock
+          labelnames:
+            - name
+            - thread_name
+            - thread_id
+        - name: instructions_per_packets
+          documentation: Instructions/packet, cycles/packet and IPC (clocks/packets)
+          namespace: vpp
+          subsystem: inst_and_clock
+          labelnames:
+            - name
+            - thread_name
+            - thread_id
+        - name: ipc
+          documentation: Instructions/packet, cycles/packet and IPC (clocks/packets)
+          namespace: vpp
+          subsystem: inst_and_clock
+          labelnames:
+            - name
+            - thread_name
+            - thread_id
+    code: |
+      perfmon reset
+      perfmon start bundle inst-and-clock
+      wait {duration}
+      perfmon stop
+      show perfmon statistics
+  - name: bundle_vpp
+    metrics:
+      gauge:
+        - name: l1_hit
+          documentation:  Cache hits and misses (L1 hit)
+          namespace: vpp
+          subsystem: cache_hierarchy
+          labelnames:
+            - name
+            - thread_name
+            - thread_id
+        - name: l1_miss
+          documentation:  Cache hits and misses (L1 miss)
+          namespace: vpp
+          subsystem: cache_hierarchy
+          labelnames:
+            - name
+            - thread_name
+            - thread_id
+        - name: l2_hit
+          documentation:  Cache hits and misses (L2 hit)
+          namespace: vpp
+          subsystem: cache_hierarchy
+          labelnames:
+            - name
+            - thread_name
+            - thread_id
+        - name: l2_miss
+          documentation:  Cache hits and misses (L2 miss)
+          namespace: vpp
+          subsystem: cache_hierarchy
+          labelnames:
+            - name
+            - thread_name
+            - thread_id
+        - name: l3_hit
+          documentation:  Cache hits and misses (L3 hit)
+          namespace: vpp
+          subsystem: cache_hierarchy
+          labelnames:
+            - name
+            - thread_name
+            - thread_id
+        - name: l3_miss
+          documentation:  Cache hits and misses (L3 miss)
+          namespace: vpp
+          subsystem: cache_hierarchy
+          labelnames:
+            - name
+            - thread_name
+            - thread_id
+    code: |
+      perfmon reset
+      perfmon start bundle cache-hierarchy
+      wait {duration}
+      perfmon stop
+      show perfmon statistics
+  - name: bundle_vpp
+    metrics:
+      gauge:
+        - name: calls
+          documentation: Load operations (calls)
+          namespace: vpp
+          subsystem: load_blocks
+          labelnames:
+            - name
+            - thread_name
+            - thread_id
+        - name: packets
+          documentation: Load operations (packets)
+          namespace: vpp
+          subsystem: load_blocks
+          labelnames:
+            - name
+            - thread_name
+            - thread_id
+    code: |
+      perfmon reset
+      perfmon start bundle load-blocks
+      wait {duration}
+      perfmon stop
+      show perfmon statistics
+  - name: bundle_vpp
+    metrics:
+      gauge:
+        - name: branches_per_call
+          documentation: Branches/call
+          namespace: vpp
+          subsystem: branch_mispred
+          labelnames:
+            - name
+            - thread_name
+            - thread_id
+        - name: branches_per_packet
+          documentation: Branches/packet
+          namespace: vpp
+          subsystem: branch_mispred
+          labelnames:
+            - name
+            - thread_name
+            - thread_id
+        - name: taken_per_call
+          documentation: Taken/call
+          namespace: vpp
+          subsystem: branch_mispred
+          labelnames:
+            - name
+            - thread_name
+            - thread_id
+        - name: taken_per_packet
+          documentation: Taken/packet
+          namespace: vpp
+          subsystem: branch_mispred
+          labelnames:
+            - name
+            - thread_name
+            - thread_id
+        - name: mis_predictions
+          documentation: Mis-predictions
+          namespace: vpp
+          subsystem: branch_mispred
+          labelnames:
+            - name
+            - thread_name
+            - thread_id
+    code: |
+      perfmon reset
+      perfmon start bundle branch-mispred
+      wait {duration}
+      perfmon stop
+      show perfmon statistics
+  - name: bundle_vpp
+    metrics:
+      gauge:
+        - name: lvl0
+          documentation: Branches/call
+          namespace: vpp
+          subsystem: power_licensing
+          labelnames:
+            - name
+            - thread_name
+            - thread_id
+        - name: lvl1
+          documentation: Branches/packet
+          namespace: vpp
+          subsystem: power_licensing
+          labelnames:
+            - name
+            - thread_name
+            - thread_id
+        - name: lvl2
+          documentation: Taken/call
+          namespace: vpp
+          subsystem: power_licensing
+          labelnames:
+            - name
+            - thread_name
+            - thread_id
+        - name: throttle
+          documentation: Taken/packet
+          namespace: vpp
+          subsystem: power_licensing
+          labelnames:
+            - name
+            - thread_name
+            - thread_id
+    code: |
+      perfmon reset
+      perfmon start bundle power-licensing
+      wait {duration}
+      perfmon stop
+      show perfmon statistics
+  - name: bundle_vpp
+    metrics:
+      gauge:
+        - name: runtime
+          documentation: RunTime
+          namespace: vpp
+          subsystem: memory_bandwidth
+          labelnames:
+            - name
+        - name: reads_mbs
+          documentation: Reads (MB/s)
+          namespace: vpp
+          subsystem: memory_bandwidth
+          labelnames:
+            - name
+        - name: writes_mbs
+          documentation: Writes (MB/s)
+          namespace: vpp
+          subsystem: memory_bandwidth
+          labelnames:
+            - name
+        - name: total_mbs
+          documentation: Total (MB/s)
+          namespace: vpp
+          subsystem: memory_bandwidth
+          labelnames:
+            - name
+    code: |
+      perfmon reset
+      perfmon start bundle memory-bandwidth
+      wait {duration}
+      perfmon stop
+      show perfmon statistics
diff --git a/resources/templates/telemetry/vpp_show_stats.yaml b/resources/templates/telemetry/vpp_show_stats.yaml
new file mode 100644
index 0000000000..16c895447c
--- /dev/null
+++ b/resources/templates/telemetry/vpp_show_stats.yaml
@@ -0,0 +1,154 @@
+---
+logging:
+  version: 1
+  formatters:
+    console:
+      format: '%(asctime)s - %(name)s - %(message)s'
+    prom:
+      format: '%(message)s'
+  handlers:
+    console:
+      class: logging.StreamHandler
+      level: INFO
+      formatter: console
+      stream: ext://sys.stdout
+    prom:
+      class: logging.handlers.RotatingFileHandler
+      level: INFO
+      formatter: prom
+      filename: /tmp/metric.prom
+      mode: w
+  loggers:
+    prom:
+      handlers: [prom]
+      level: INFO
+      propagate: False
+  root:
+    level: INFO
+    handlers: [console]
+scheduler:
+  duration: 1
+programs:
+  - name: bundle_vpp
+    metrics:
+      counter:
+        - name: calls
+          documentation: Number of calls total
+          labelnames:
+            - name
+            - state
+            - thread_name
+            - thread_id
+            - thread_lcore
+        - name: vectors
+          documentation: Number of vectors total
+          labelnames:
+            - name
+            - state
+            - thread_name
+            - thread_id
+            - thread_lcore
+        - name: suspends
+          documentation: Number of suspends total
+          labelnames:
+            - name
+            - state
+            - thread_name
+            - thread_id
+            - thread_lcore
+        - name: clocks
+          documentation: Number of clocks total
+          labelnames:
+            - name
+            - state
+            - thread_name
+            - thread_id
+            - thread_lcore
+        - name: vectors_calls
+          documentation: Number of vectors per call
+          labelnames:
+            - name
+            - state
+            - thread_name
+            - thread_id
+            - thread_lcore
+    code: |
+      show runtime
+  - name: bundle_vpp
+    metrics:
+      counter:
+        - name: rx_packets
+          documentation: Number of received packets for interface
+          labelnames:
+            - name
+            - index
+        - name: rx_bytes
+          documentation: Number of received bytes for interface
+          labelnames:
+            - name
+            - index
+        - name: rx_error
+          documentation: Number of errors on interface
+          labelnames:
+            - name
+            - index
+        - name: tx_packets
+          documentation: Number of transitted packets for interface
+          labelnames:
+            - name
+            - index
+        - name: tx_bytes
+          documentation: Number of transitted bytes for interface
+          labelnames:
+            - name
+            - index
+        - name: tx_error
+          documentation: Number of errors on interface
+          labelnames:
+            - name
+            - index
+        - name: drops
+          documentation: Number of dropped packets for interface
+          labelnames:
+            - name
+            - index
+        - name: punt
+          documentation: Number of punted packets for interface
+          labelnames:
+            - name
+            - index
+        - name: ip4
+          documentation: Number of IPv4 packets for interface
+          labelnames:
+            - name
+            - index
+        - name: ip6
+          documentation: Number of IPv6 packets for interface
+          labelnames:
+            - name
+            - index
+        - name: rx_no_buf
+          documentation: Number of out of buffer RX packets on interface
+          labelnames:
+            - name
+            - index
+        - name: rx_miss
+          documentation: Number of missed RX packets on interface
+          labelnames:
+            - name
+            - index
+    code: |
+      show interface
+  - name: bundle_vpp
+    metrics:
+      counter:
+        - name: node_counter
+          documentation: Node counter
+          labelnames:
+            - name
+            - reason
+            - severity
+            - thread_name
+            - thread_id
+    code: |
+      show node counters verbose