consul/test/integration/connect/envoy/helpers.bash

256 lines
6.4 KiB
Bash
Raw Normal View History

#!/bin/bash
# retry based on
# https://github.com/fernandoacorreia/azure-docker-registry/blob/master/tools/scripts/create-registry-server
# under MIT license.
function retry {
local n=1
local max=$1
shift
local delay=$1
shift
while true; do
"$@" && break || {
exit=$?
if [[ $n -lt $max ]]; then
((n++))
echo "Command failed. Attempt $n/$max:"
sleep $delay;
else
echo "The command has failed after $n attempts." >&2
return $exit
fi
}
done
}
function retry_default {
retry 5 1 $@
}
function echored {
tput setaf 1
tput bold
echo $@
tput sgr0
}
function echogreen {
tput setaf 2
tput bold
echo $@
tput sgr0
}
function echoyellow {
tput setaf 3
tput bold
echo $@
tput sgr0
}
function echoblue {
tput setaf 4
tput bold
echo $@
tput sgr0
}
function get_cert {
local HOSTPORT=$1
openssl s_client -connect $HOSTPORT \
-showcerts 2>/dev/null \
| openssl x509 -noout -text
}
function assert_proxy_presents_cert_uri {
local HOSTPORT=$1
local SERVICENAME=$2
CERT=$(retry_default get_cert $HOSTPORT)
echo "WANT SERVICE: $SERVICENAME"
echo "GOT CERT:"
echo "$CERT"
echo "$CERT" | grep -Eo "URI:spiffe://([a-zA-Z0-9-]+).consul/ns/default/dc/dc1/svc/$SERVICENAME"
}
function assert_envoy_version {
local ADMINPORT=$1
run retry_default curl -f -s localhost:$ADMINPORT/server_info
[ "$status" -eq 0 ]
# Envoy 1.8.0 returns a plain text line like
# envoy 5d25f466c3410c0dfa735d7d4358beb76b2da507/1.8.0/Clean/DEBUG live 3 3 0
# Later versions return JSON.
if (echo $output | grep '^envoy') ; then
VERSION=$(echo $output | cut -d ' ' -f 2)
else
VERSION=$(echo $output | jq -r '.version')
fi
echo "Status=$status"
echo "Output=$output"
echo "---"
echo "Got version=$VERSION"
echo "Want version=$ENVOY_VERSION"
echo $VERSION | grep "/$ENVOY_VERSION/"
}
function get_envoy_listener_filters {
local HOSTPORT=$1
run retry_default curl -s -f $HOSTPORT/config_dump
[ "$status" -eq 0 ]
echo "$output" | jq --raw-output '.configs[2].dynamic_active_listeners[].listener | "\(.name) \( .filter_chains[0].filters | map(.name) | join(","))"'
}
function get_envoy_stats_flush_interval {
local HOSTPORT=$1
run retry_default curl -s -f $HOSTPORT/config_dump
[ "$status" -eq 0 ]
#echo "$output" > /workdir/s1_envoy_dump.json
echo "$output" | jq --raw-output '.configs[0].bootstrap.stats_flush_interval'
}
# snapshot_envoy_admin is meant to be used from a teardown scriptlet from the host.
function snapshot_envoy_admin {
local HOSTPORT=$1
local ENVOY_NAME=$2
docker_wget "http://${HOSTPORT}/config_dump" -q -O - > "./workdir/envoy/${ENVOY_NAME}-config_dump.json"
docker_wget "http://${HOSTPORT}/clusters?format=json" -q -O - > "./workdir/envoy/${ENVOY_NAME}-clusters.json"
}
function get_healthy_upstream_endpoint_count {
local HOSTPORT=$1
local CLUSTER_NAME=$2
run retry_default curl -s -f "http://${HOSTPORT}/clusters?format=json"
[ "$status" -eq 0 ]
echo "$output" | jq --raw-output "
.cluster_statuses[]
| select(.name|startswith(\"${CLUSTER_NAME}.default.dc1.internal.\"))
| .host_statuses[].health_status
| select(.eds_health_status == \"HEALTHY\")
| length"
}
function assert_upstream_has_healthy_endpoints_once {
local HOSTPORT=$1
local CLUSTER_NAME=$2
local EXPECT_COUNT=$3
GOT_COUNT=$(get_healthy_upstream_endpoint_count $HOSTPORT $CLUSTER_NAME)
[ "$GOT_COUNT" -eq $EXPECT_COUNT ]
}
function assert_upstream_has_healthy_endpoints {
local HOSTPORT=$1
local CLUSTER_NAME=$2
local EXPECT_COUNT=$3
run retry 30 1 assert_upstream_has_healthy_endpoints_once $HOSTPORT $CLUSTER_NAME $EXPECT_COUNT
[ "$status" -eq 0 ]
}
test: for envoy integration tests, wait until 's2' is healthy in consul before interrogating envoy (#6108) When the envoy healthy panic threshold was explicitly disabled as part of L7 traffic management it changed how envoy decided to load balance to endpoints in a cluster. This only matters when envoy is in "panic mode" aka "when you have a bunch of unhealthy endpoints". Panic mode sends traffic to unhealthy instances in certain circumstances. Note: Prior to explicitly disabling the healthy panic threshold, the default value is 50%. What was happening is that the test harness was bringing up consul the sidecars, and the service instances all at once and sometimes the proxies wouldn't have time to be checked by consul to be labeled as 'passing' in the catalog before a round of EDS happened. The xDS server in consul effectively queries /v1/health/connect/s2 and gets 1 result, but that one result has a 'critical' check so the xDS server sends back that endpoint labeled as UNHEALTHY. Envoy sees that 100% of the endpoints in the cluster are unhealthy and would enter panic mode and still send traffic to s2. This is why the test suites PRIOR to disabling the healthy panic threshold worked. They were _incorrectly_ passing. When the healthy panic threshol is disabled, envoy never enters panic mode in this situation and thus the cluster has zero healthy endpoints so load balancing goes nowhere and the tests fail. Why does this only affect the test suites for envoy 1.8.0? My guess is that https://github.com/envoyproxy/envoy/pull/4442 was merged into the 1.9.x series and somehow that plays a role. This PR modifies the bats scripts to explicitly wait until the upstream sidecar is healthy as measured by /v1/health/connect/s2?passing BEFORE trying to interrogate envoy which should make the tests less racy.
2019-07-10 20:58:25 +00:00
function get_healthy_service_count {
local SERVICE_NAME=$1
run retry_default curl -s -f 127.0.0.1:8500/v1/health/connect/${SERVICE_NAME}?passing
[ "$status" -eq 0 ]
echo "$output" | jq --raw-output '. | length'
}
function assert_service_has_healthy_instances_once {
test: for envoy integration tests, wait until 's2' is healthy in consul before interrogating envoy (#6108) When the envoy healthy panic threshold was explicitly disabled as part of L7 traffic management it changed how envoy decided to load balance to endpoints in a cluster. This only matters when envoy is in "panic mode" aka "when you have a bunch of unhealthy endpoints". Panic mode sends traffic to unhealthy instances in certain circumstances. Note: Prior to explicitly disabling the healthy panic threshold, the default value is 50%. What was happening is that the test harness was bringing up consul the sidecars, and the service instances all at once and sometimes the proxies wouldn't have time to be checked by consul to be labeled as 'passing' in the catalog before a round of EDS happened. The xDS server in consul effectively queries /v1/health/connect/s2 and gets 1 result, but that one result has a 'critical' check so the xDS server sends back that endpoint labeled as UNHEALTHY. Envoy sees that 100% of the endpoints in the cluster are unhealthy and would enter panic mode and still send traffic to s2. This is why the test suites PRIOR to disabling the healthy panic threshold worked. They were _incorrectly_ passing. When the healthy panic threshol is disabled, envoy never enters panic mode in this situation and thus the cluster has zero healthy endpoints so load balancing goes nowhere and the tests fail. Why does this only affect the test suites for envoy 1.8.0? My guess is that https://github.com/envoyproxy/envoy/pull/4442 was merged into the 1.9.x series and somehow that plays a role. This PR modifies the bats scripts to explicitly wait until the upstream sidecar is healthy as measured by /v1/health/connect/s2?passing BEFORE trying to interrogate envoy which should make the tests less racy.
2019-07-10 20:58:25 +00:00
local SERVICE_NAME=$1
local EXPECT_COUNT=$2
GOT_COUNT=$(get_healthy_service_count $SERVICE_NAME)
[ "$GOT_COUNT" -eq $EXPECT_COUNT ]
}
function assert_service_has_healthy_instances {
local SERVICE_NAME=$1
local EXPECT_COUNT=$2
run retry 30 1 assert_service_has_healthy_instances_once $SERVICE_NAME $EXPECT_COUNT
test: for envoy integration tests, wait until 's2' is healthy in consul before interrogating envoy (#6108) When the envoy healthy panic threshold was explicitly disabled as part of L7 traffic management it changed how envoy decided to load balance to endpoints in a cluster. This only matters when envoy is in "panic mode" aka "when you have a bunch of unhealthy endpoints". Panic mode sends traffic to unhealthy instances in certain circumstances. Note: Prior to explicitly disabling the healthy panic threshold, the default value is 50%. What was happening is that the test harness was bringing up consul the sidecars, and the service instances all at once and sometimes the proxies wouldn't have time to be checked by consul to be labeled as 'passing' in the catalog before a round of EDS happened. The xDS server in consul effectively queries /v1/health/connect/s2 and gets 1 result, but that one result has a 'critical' check so the xDS server sends back that endpoint labeled as UNHEALTHY. Envoy sees that 100% of the endpoints in the cluster are unhealthy and would enter panic mode and still send traffic to s2. This is why the test suites PRIOR to disabling the healthy panic threshold worked. They were _incorrectly_ passing. When the healthy panic threshol is disabled, envoy never enters panic mode in this situation and thus the cluster has zero healthy endpoints so load balancing goes nowhere and the tests fail. Why does this only affect the test suites for envoy 1.8.0? My guess is that https://github.com/envoyproxy/envoy/pull/4442 was merged into the 1.9.x series and somehow that plays a role. This PR modifies the bats scripts to explicitly wait until the upstream sidecar is healthy as measured by /v1/health/connect/s2?passing BEFORE trying to interrogate envoy which should make the tests less racy.
2019-07-10 20:58:25 +00:00
[ "$status" -eq 0 ]
}
function docker_consul {
docker run -ti --rm --network container:envoy_consul_1 consul-dev $@
}
function docker_wget {
docker run -ti --rm --network container:envoy_consul_1 alpine:3.9 wget $@
}
function must_match_in_statsd_logs {
run cat /workdir/statsd/statsd.log
COUNT=$( echo "$output" | grep -Ec $1 )
echo "COUNT of '$1' matches: $COUNT"
[ "$status" == 0 ]
[ "$COUNT" -gt "0" ]
}
function must_match_in_prometheus_response {
run curl -f -s $1/metrics
COUNT=$( echo "$output" | grep -Ec $2 )
echo "OUTPUT head -n 10"
echo "$output" | head -n 10
echo "COUNT of '$2' matches: $COUNT"
[ "$status" == 0 ]
[ "$COUNT" -gt "0" ]
}
# must_fail_tcp_connection checks that a request made through an upstream fails,
# probably due to authz being denied if all other tests passed already. Although
# we are using curl, this only works as expected for TCP upstreams as we are
# checking TCP-level errors. HTTP upstreams will return a valid 503 generated by
# Envoy rather than a connection-level error.
function must_fail_tcp_connection {
# Attempt to curl through upstream
run curl -s -v -f -d hello $1
echo "OUTPUT $output"
# Should fail during handshake and return "got nothing" error
[ "$status" == "52" ]
# Verbose output should enclude empty reply
echo "$output" | grep 'Empty reply from server'
}
# must_fail_http_connection see must_fail_tcp_connection but this expects Envoy
# to generate a 503 response since the upstreams have refused connection.
function must_fail_http_connection {
# Attempt to curl through upstream
run curl -s -i -d hello $1
echo "OUTPUT $output"
# Should fail request with 503
echo "$output" | grep '503 Service Unavailable'
}
function gen_envoy_bootstrap {
SERVICE=$1
ADMIN_PORT=$2
if output=$(docker_consul connect envoy -bootstrap \
-proxy-id $SERVICE-sidecar-proxy \
-admin-bind 0.0.0.0:$ADMIN_PORT 2>&1); then
# All OK, write config to file
echo "$output" > workdir/envoy/$SERVICE-bootstrap.json
else
status=$?
# Command failed, instead of swallowing error (printed on stdout by docker
# it seems) by writing it to file, echo it
echo "$output"
return $status
fi
}