[telemetry] enable metrics and on networkpolicy

Closes: aptos-labs#949

crates/aptos-telemetry/Cargo.toml

@@ -10,6 +10,7 @@ publish = false
 edition = "2018"
 
 [dependencies]
+once_cell = "1.10.0"
 reqwest = { version = "0.11.10", features = ["json"] }
 serde = { version = "1.0.137", features = ["derive"], default-features = false }
 serde_json = "1.0.81"

crates/aptos-telemetry/src/constants.rs

@@ -9,8 +9,10 @@
 pub const APTOS_GA_MEASUREMENT_ID: &str = "G-ZX4L6WPCFZ";
 pub const APTOS_GA_API_SECRET: &str = "ArtslKPTTjeiMi1n-IR39g";
 
-pub const HTTPBIN_URL: &str = "http://httpbin.org/ip";
-pub const GA4_URL: &str = "http://www.google-analytics.com/mp/collect";
+pub const HTTPBIN_URL: &str = "https://httpbin.org/ip";
+// measurement protocol requires HTTPS
+// https://developers.google.com/analytics/devguides/collection/protocol/v1/reference#transport
+pub const GA4_URL: &str = "https://www.google-analytics.com/mp/collect";
 
 // Timeouts
 pub const NETWORK_PUSH_TIME_SECS: u64 = 30;

crates/aptos-telemetry/src/lib.rs

@@ -7,7 +7,8 @@
 pub mod constants;
 
 use aptos_logger::prelude::*;
-use aptos_metrics::json_metrics::get_git_rev;
+use aptos_metrics::{json_metrics::get_git_rev, register_int_counter_vec, IntCounterVec};
+use once_cell::sync::Lazy;
 use serde::{Deserialize, Serialize};
 use std::{
     collections::HashMap,
@@ -20,6 +21,24 @@ pub const GA_MEASUREMENT_ID: &str = "GA_MEASUREMENT_ID";
 pub const GA_API_SECRET: &str = "GA_API_SECRET";
 pub const APTOS_TELEMETRY_DISABLE: &str = "APTOS_TELEMETRY_DISABLE";
 
+pub static APTOS_TELEMETRY_SUCCESS: Lazy<IntCounterVec> = Lazy::new(|| {
+    register_int_counter_vec!(
+        "aptos_telemetry_success",
+        "Number of telemetry events successfully sent",
+        &["event_name"]
+    )
+    .unwrap()
+});
+
+pub static APTOS_TELEMETRY_FAILURE: Lazy<IntCounterVec> = Lazy::new(|| {
+    register_int_counter_vec!(
+        "aptos_telemetry_failure",
+        "Number of telemetry events failed to send",
+        &["event_name"]
+    )
+    .unwrap()
+});
+
 #[derive(Debug, Serialize, Deserialize)]
 struct MetricsDump {
     client_id: String,
@@ -86,7 +105,7 @@ pub async fn send_data(event_name: String, user_id: String, event_params: HashMa
         .unwrap_or_else(|_| constants::APTOS_GA_MEASUREMENT_ID.to_string());
 
     let metrics_event = MetricsEvent {
-        name: event_name,
+        name: event_name.clone(),
         params: event_params,
     };
 
@@ -114,8 +133,33 @@ pub async fn send_data(event_name: String, user_id: String, event_params: HashMa
             .send()
             .await;
         match res {
-            Ok(_) => debug!("Sent telemetry data {:?}", &metrics_dump),
-            Err(e) => debug!("{:?}", e),
+            Ok(res) => {
+                let status_code = res.status().as_u16();
+                if status_code > 200 && status_code < 299 {
+                    info!("Sent telemetry event {}", event_name.as_str());
+                    debug!("Sent telemetry data {:?}", &metrics_dump);
+                    APTOS_TELEMETRY_SUCCESS
+                        .with_label_values(&[event_name.as_str()])
+                        .inc();
+                } else {
+                    info!(
+                        "Failed to send telemetry event {}: {}",
+                        res.status(),
+                        event_name.as_str()
+                    );
+                    debug!("{:?}", res.text().await);
+                    APTOS_TELEMETRY_FAILURE
+                        .with_label_values(&[event_name.as_str()])
+                        .inc();
+                }
+            }
+            Err(e) => {
+                info!("Failed to send telemetry event {}", event_name.as_str());
+                debug!("{:?}", e);
+                APTOS_TELEMETRY_FAILURE
+                    .with_label_values(&[event_name.as_str()])
+                    .inc();
+            }
         }
     });
 }

docker/faucet/Dockerfile

@@ -24,7 +24,7 @@ RUN IMAGE_TARGETS="test" ./docker/build-common.sh
 ### Production Image ###
 FROM debian-base AS pre-test
 
-RUN apt-get update && apt-get install -y libssl1.1 nano net-tools tcpdump iproute2 netcat \
+RUN apt-get update && apt-get install -y libssl1.1 ca-certificates nano net-tools tcpdump iproute2 netcat \
     && apt-get clean && rm -r /var/lib/apt/lists/*
 
 RUN mkdir -p /opt/aptos/bin  /aptos/client/data/wallet/

docker/forge/Dockerfile

@@ -23,7 +23,7 @@ RUN IMAGE_TARGETS="test" ./docker/build-common.sh
 
 FROM debian-base
 
-RUN apt-get update && apt-get install -y libssl1.1 openssh-client wget busybox git unzip awscli && apt-get clean && rm -r /var/lib/apt/lists/*
+RUN apt-get update && apt-get install -y libssl1.1 ca-certificates openssh-client wget busybox git unzip awscli && apt-get clean && rm -r /var/lib/apt/lists/*
 
 RUN mkdir /aptos
 COPY rust-toolchain /aptos/rust-toolchain

docker/indexer/Dockerfile

@@ -24,7 +24,7 @@ RUN IMAGE_TARGETS="release" ./docker/build-common.sh
 ### Production Image ###
 FROM debian-base AS pre-prod
 
-RUN apt-get update && apt-get install -y libssl1.1 net-tools tcpdump iproute2 netcat libpq-dev \
+RUN apt-get update && apt-get install -y libssl1.1 ca-certificates net-tools tcpdump iproute2 netcat libpq-dev \
     && apt-get clean && rm -r /var/lib/apt/lists/*
 
 RUN mkdir -p /opt/aptos/bin

docker/init/Dockerfile

@@ -24,7 +24,7 @@ RUN IMAGE_TARGETS="release" ./docker/build-common.sh
 ### Production Image ###
 FROM debian-base AS pre-prod
 
-RUN apt-get update && apt-get -y install libssl1.1 wget busybox gettext-base && apt-get clean && rm -r /var/lib/apt/lists/*
+RUN apt-get update && apt-get -y install libssl1.1 ca-certificates wget busybox gettext-base && apt-get clean && rm -r /var/lib/apt/lists/*
 RUN cd /usr/local/bin && wget "https://storage.googleapis.com/kubernetes-release/release/v1.18.6/bin/linux/amd64/kubectl" -O kubectl && chmod +x kubectl
 RUN cd /usr/local/bin && wget "https://releases.hashicorp.com/vault/1.5.0/vault_1.5.0_linux_amd64.zip" -O- | busybox unzip - && chmod +x vault
 

docker/safety-rules/Dockerfile

@@ -24,7 +24,7 @@ RUN IMAGE_TARGETS="release" ./docker/build-common.sh
 ### Production Image ###
 FROM debian-base AS prod
 
-RUN apt-get update && apt-get install -y libssl1.1 && apt-get clean && rm -r /var/lib/apt/lists/*
+RUN apt-get update && apt-get install -y libssl1.1 ca-certificates && apt-get clean && rm -r /var/lib/apt/lists/*
 
 RUN addgroup --system --gid 6180 aptos && adduser --system --ingroup aptos --no-create-home --uid 6180 aptos
 

docker/txn-emitter/Dockerfile

@@ -24,7 +24,7 @@ RUN IMAGE_TARGETS="release" ./docker/build-common.sh
 ### Production Image ###
 FROM debian-base AS pre-prod
 
-RUN apt-get update && apt-get -y install libssl1.1 wget busybox gettext-base && apt-get clean && rm -r /var/lib/apt/lists/*
+RUN apt-get update && apt-get -y install libssl1.1 ca-certificates wget busybox gettext-base && apt-get clean && rm -r /var/lib/apt/lists/*
 
 RUN mkdir -p /opt/aptos/bin
 COPY --from=builder /aptos/target/release/transaction-emitter /usr/local/bin

docker/validator/Dockerfile

@@ -24,7 +24,7 @@ RUN IMAGE_TARGETS="release" ./docker/build-common.sh
 ### Production Image ###
 FROM debian-base AS prod
 
-RUN apt-get update && apt-get install -y libssl1.1 && apt-get clean && rm -r /var/lib/apt/lists/*
+RUN apt-get update && apt-get install -y libssl1.1 ca-certificates && apt-get clean && rm -r /var/lib/apt/lists/*
 
 ### Needed to run debugging tools like perf
 RUN apt-get update && apt-get install -y linux-tools-4.19 sudo procps

terraform/helm/aptos-node/templates/validator.yaml

@@ -198,13 +198,14 @@ spec:
   - ports:
     - protocol: TCP
       port: 6180
-  # Aptos telemetry
-  - ports:
-    - protocol: TCP
-      port: 80
   # DNS
   - to:
     - namespaceSelector: {}
     ports:
     - protocol: UDP
       port: 53
+  # Enable HTTPS telemetry
+  - ports:
+    - protocol: TCP
+      port: 443
+

terraform/helm/validator/templates/validator.yaml

@@ -297,3 +297,7 @@ spec:
     ports:
     - protocol: UDP
       port: 53
+  # Enable HTTPS telemetry
+  - ports:
+    - protocol: TCP
+      port: 443