Add support to trigger testnet from a PR (#1434)

* Add support for different node counts

* Update variable names

* Delete network even after failures

* Add array for node counts

* Changed number of nodes to a space separated string of numbers

* Adjust number of nodes

* Snap will not be published if the env variable DO_NOT_PUBLISH_SNAP is set

* Address review comments

* Replaced influx db URL

ci/snap.sh

@@ -57,5 +57,7 @@ snapcraft
 source ci/upload_ci_artifact.sh
 upload_ci_artifact solana_*.snap
 
-echo --- publish: $CHANNEL channel
-$DRYRUN snapcraft push solana_*.snap --release $CHANNEL
+if [[ -z $DO_NOT_PUBLISH_SNAP ]]; then
+  echo --- publish: $CHANNEL channel
+  $DRYRUN snapcraft push solana_*.snap --release $CHANNEL
+fi

ci/solana-testnet.yml

@@ -0,0 +1,18 @@
+steps:
+  - command: "ci/snap.sh"
+    label: "create snap"
+
+  - wait
+
+  - command: "ci/testnet-automation.sh"
+    label: "run testnet"
+    agents:
+      - "queue=metrics"
+
+  - wait: ~
+    continue_on_failure: true
+
+  - command: "ci/testnet-automation-cleanup.sh"
+    label: "delete testnet"
+    agents:
+      - "queue=metrics"

ci/testnet-automation-cleanup.sh

@@ -0,0 +1,9 @@
+#!/bin/bash -e
+
+cd "$(dirname "$0")/.."
+
+echo --- find testnet configuration
+net/gce.sh config -p testnet-automation
+
+echo --- delete testnet
+net/gce.sh delete -p testnet-automation

ci/testnet-automation.sh

@@ -0,0 +1,75 @@
+#!/bin/bash -e
+
+cd "$(dirname "$0")/.."
+
+echo --- downloading snap from build artifacts
+buildkite-agent artifact download "solana_*.snap" .
+
+# shellcheck disable=SC1091
+source ci/upload_ci_artifact.sh
+
+[[ -n $ITERATION_WAIT ]] || ITERATION_WAIT=300
+[[ -n $NUMBER_OF_NODES ]] || NUMBER_OF_NODES="10 25 50 100"
+
+launchTestnet() {
+  declare nodeCount=$1
+  echo --- setup "$nodeCount" node test
+  net/gce.sh create \
+    -n "$nodeCount" -c 2 \
+    -G "n1-standard-16 --accelerator count=2,type=nvidia-tesla-v100" \
+    -p testnet-automation -z us-west1-b
+
+  echo --- configure database
+  net/init-metrics.sh -e
+
+  echo --- start "$nodeCount" node test
+  net/net.sh start -o noValidatorSanity -S solana_*.snap
+
+  echo --- wait "$ITERATION_WAIT" seconds to complete test
+  sleep "$ITERATION_WAIT"
+
+  declare q_mean_tps='
+    SELECT round(mean("sum_count")) FROM (
+      SELECT sum("count") AS "sum_count"
+        FROM "testnet-automation"."autogen"."counter-banking_stage-process_transactions"
+        WHERE time > now() - 300s GROUP BY time(1s)
+    )'
+
+  declare q_max_tps='
+    SELECT round(max("sum_count")) FROM (
+      SELECT sum("count") AS "sum_count"
+        FROM "testnet-automation"."autogen"."counter-banking_stage-process_transactions"
+        WHERE time > now() - 300s GROUP BY time(1s)
+    )'
+
+  declare q_mean_finality='
+    SELECT round(mean("duration_ms"))
+      FROM "testnet-automation"."autogen"."leader-finality"
+      WHERE time > now() - 300s'
+
+  declare q_max_finality='
+    SELECT round(max("duration_ms"))
+      FROM "testnet-automation"."autogen"."leader-finality"
+      WHERE time > now() - 300s'
+
+  declare q_99th_finality='
+    SELECT round(percentile("duration_ms", 99))
+      FROM "testnet-automation"."autogen"."leader-finality"
+      WHERE time > now() - 300s'
+
+  curl -G "https://metrics.solana.com:8086/query?u=${INFLUX_USERNAME}&p=${INFLUX_PASSWORD}" \
+    --data-urlencode "db=$INFLUX_DATABASE" \
+    --data-urlencode "q=$q_mean_tps;$q_max_tps;$q_mean_finality;$q_max_finality;$q_99th_finality" \
+    >>TPS"$nodeCount".log
+
+  upload_ci_artifact TPS"$nodeCount".log
+}
+
+# This is needed, because buildkite doesn't let us define an array of numbers.
+# The array is defined as a space separated string of numbers
+# shellcheck disable=SC2206
+nodes_count_array=($NUMBER_OF_NODES)
+
+for n in "${nodes_count_array[@]}"; do
+  launchTestnet "$n"
+done

net/net.sh

@@ -159,7 +159,7 @@ startValidator() {
   declare ipAddress=$1
   declare logFile="$netLogDir/validator-$ipAddress.log"
 
-  echo "--- Starting validator: $leaderIp"
+  echo "--- Starting validator: $ipAddress"
   echo "start log: $logFile"
   (
     set -x