debian-mirror-gitlab/spec/lib/gitlab/usage_data/topology_spec.rb
2020-08-09 17:41:57 +05:30

562 lines
18 KiB
Ruby

# frozen_string_literal: true
require 'spec_helper'
RSpec.describe Gitlab::UsageData::Topology do
include UsageDataHelpers
describe '#topology_usage_data' do
subject { described_class.new.topology_usage_data }
before do
# this pins down time shifts when benchmarking durations
allow(Process).to receive(:clock_gettime).and_return(0)
end
context 'when embedded Prometheus server is enabled' do
before do
expect(Gitlab::Prometheus::Internal).to receive(:prometheus_enabled?).and_return(true)
expect(Gitlab::Prometheus::Internal).to receive(:uri).and_return('http://prom:9090')
end
context 'tracking node metrics' do
it 'contains node level metrics for each instance' do
expect_prometheus_api_to(
receive_app_request_volume_query,
receive_node_memory_query,
receive_node_cpu_count_query,
receive_node_uname_info_query,
receive_node_service_memory_rss_query,
receive_node_service_memory_uss_query,
receive_node_service_memory_pss_query,
receive_node_service_process_count_query,
receive_node_service_app_server_workers_query
)
expect(subject[:topology]).to eq({
duration_s: 0,
application_requests_per_hour: 36,
failures: [],
nodes: [
{
node_memory_total_bytes: 512,
node_cpus: 8,
node_uname_info: {
machine: 'x86_64',
sysname: 'Linux',
release: '4.19.76-linuxkit'
},
node_services: [
{
name: 'web',
process_count: 10,
process_memory_rss: 300,
process_memory_uss: 301,
process_memory_pss: 302,
server: 'puma'
},
{
name: 'sidekiq',
process_count: 5,
process_memory_rss: 303
}
]
},
{
node_memory_total_bytes: 1024,
node_cpus: 16,
node_uname_info: {
machine: 'x86_64',
sysname: 'Linux',
release: '4.15.0-101-generic'
},
node_services: [
{
name: 'sidekiq',
process_count: 15,
process_memory_rss: 400,
process_memory_pss: 401
},
{
name: 'redis',
process_count: 1,
process_memory_rss: 402
},
{
name: 'registry',
process_count: 1
},
{
name: 'web',
server: 'unicorn'
}
]
}
]
})
end
end
context 'and some node memory metrics are missing' do
it 'removes the respective entries and includes the failures' do
expect_prometheus_api_to(
receive_app_request_volume_query(result: []),
receive_node_memory_query(result: []),
receive_node_cpu_count_query,
receive_node_uname_info_query,
receive_node_service_memory_rss_query(result: []),
receive_node_service_memory_uss_query(result: []),
receive_node_service_memory_pss_query,
receive_node_service_process_count_query,
receive_node_service_app_server_workers_query(result: [])
)
expect(subject[:topology]).to eq({
duration_s: 0,
failures: [
{ 'app_requests' => 'empty_result' },
{ 'node_memory' => 'empty_result' },
{ 'service_rss' => 'empty_result' },
{ 'service_uss' => 'empty_result' },
{ 'service_workers' => 'empty_result' }
],
nodes: [
{
node_cpus: 16,
node_uname_info: {
machine: 'x86_64',
release: '4.15.0-101-generic',
sysname: 'Linux'
},
node_services: [
{
name: 'sidekiq',
process_count: 15,
process_memory_pss: 401
},
{
name: 'redis',
process_count: 1
},
{
name: 'registry',
process_count: 1
}
]
},
{
node_cpus: 8,
node_uname_info: {
machine: 'x86_64',
release: '4.19.76-linuxkit',
sysname: 'Linux'
},
node_services: [
{
name: 'web',
process_count: 10,
process_memory_pss: 302
},
{
name: 'sidekiq',
process_count: 5
}
]
}
]
})
end
end
context 'and services run on the same node but report different instance values' do
let(:node_memory_response) do
[
{
'metric' => { 'instance' => 'localhost:9100' },
'value' => [1000, '512']
}
]
end
let(:node_uname_info_response) do
[
{
"metric" => {
"__name__" => "node_uname_info",
"domainname" => "(none)",
"instance" => "127.0.0.1:9100",
"job" => "node_exporter",
"machine" => "x86_64",
"nodename" => "127.0.0.1",
"release" => "4.19.76-linuxkit",
"sysname" => "Linux"
},
"value" => [1592463033.359, "1"]
}
]
end
# The services in this response should all be mapped to localhost i.e. the same node
let(:service_memory_response) do
[
{
'metric' => { 'instance' => 'localhost:8080', 'job' => 'gitlab-rails' },
'value' => [1000, '10']
},
{
'metric' => { 'instance' => '127.0.0.1:8090', 'job' => 'gitlab-sidekiq' },
'value' => [1000, '11']
},
{
'metric' => { 'instance' => '0.0.0.0:9090', 'job' => 'prometheus' },
'value' => [1000, '12']
},
{
'metric' => { 'instance' => '[::1]:1234', 'job' => 'redis' },
'value' => [1000, '13']
},
{
'metric' => { 'instance' => '[::]:1234', 'job' => 'postgres' },
'value' => [1000, '14']
}
]
end
it 'normalizes equivalent instance values and maps them to the same node' do
expect_prometheus_api_to(
receive_app_request_volume_query(result: []),
receive_node_memory_query(result: node_memory_response),
receive_node_cpu_count_query(result: []),
receive_node_uname_info_query(result: node_uname_info_response),
receive_node_service_memory_rss_query(result: service_memory_response),
receive_node_service_memory_uss_query(result: []),
receive_node_service_memory_pss_query(result: []),
receive_node_service_process_count_query(result: []),
receive_node_service_app_server_workers_query(result: [])
)
expect(subject[:topology]).to eq({
duration_s: 0,
failures: [
{ 'app_requests' => 'empty_result' },
{ 'node_cpus' => 'empty_result' },
{ 'service_uss' => 'empty_result' },
{ 'service_pss' => 'empty_result' },
{ 'service_process_count' => 'empty_result' },
{ 'service_workers' => 'empty_result' }
],
nodes: [
{
node_memory_total_bytes: 512,
node_uname_info: {
machine: 'x86_64',
sysname: 'Linux',
release: '4.19.76-linuxkit'
},
node_services: [
{
name: 'web',
process_memory_rss: 10
},
{
name: 'sidekiq',
process_memory_rss: 11
},
{
name: 'prometheus',
process_memory_rss: 12
},
{
name: 'redis',
process_memory_rss: 13
},
{
name: 'postgres',
process_memory_rss: 14
}
]
}
]
})
end
end
context 'and node metrics are missing but service metrics exist' do
it 'still reports service metrics' do
expect_prometheus_api_to(
receive_app_request_volume_query(result: []),
receive_node_memory_query(result: []),
receive_node_cpu_count_query(result: []),
receive_node_uname_info_query(result: []),
receive_node_service_memory_rss_query,
receive_node_service_memory_uss_query(result: []),
receive_node_service_memory_pss_query(result: []),
receive_node_service_process_count_query(result: []),
receive_node_service_app_server_workers_query(result: [])
)
expect(subject[:topology]).to eq({
duration_s: 0,
failures: [
{ 'app_requests' => 'empty_result' },
{ 'node_memory' => 'empty_result' },
{ 'node_cpus' => 'empty_result' },
{ 'node_uname_info' => 'empty_result' },
{ 'service_uss' => 'empty_result' },
{ 'service_pss' => 'empty_result' },
{ 'service_process_count' => 'empty_result' },
{ 'service_workers' => 'empty_result' }
],
nodes: [
{
node_services: [
{
name: 'web',
process_memory_rss: 300
},
{
name: 'sidekiq',
process_memory_rss: 303
}
]
},
{
node_services: [
{
name: 'sidekiq',
process_memory_rss: 400
},
{
name: 'redis',
process_memory_rss: 402
}
]
}
]
})
end
end
context 'and an error is raised when querying Prometheus' do
it 'returns empty result with failures' do
expect_prometheus_api_to receive(:query)
.at_least(:once)
.and_raise(Gitlab::PrometheusClient::ConnectionError)
expect(subject[:topology]).to eq({
duration_s: 0,
failures: [
{ 'app_requests' => 'Gitlab::PrometheusClient::ConnectionError' },
{ 'node_memory' => 'Gitlab::PrometheusClient::ConnectionError' },
{ 'node_cpus' => 'Gitlab::PrometheusClient::ConnectionError' },
{ 'node_uname_info' => 'Gitlab::PrometheusClient::ConnectionError' },
{ 'service_rss' => 'Gitlab::PrometheusClient::ConnectionError' },
{ 'service_uss' => 'Gitlab::PrometheusClient::ConnectionError' },
{ 'service_pss' => 'Gitlab::PrometheusClient::ConnectionError' },
{ 'service_process_count' => 'Gitlab::PrometheusClient::ConnectionError' },
{ 'service_workers' => 'Gitlab::PrometheusClient::ConnectionError' }
],
nodes: []
})
end
end
end
context 'when embedded Prometheus server is disabled' do
it 'returns empty result with no failures' do
expect(Gitlab::Prometheus::Internal).to receive(:prometheus_enabled?).and_return(false)
expect(subject[:topology]).to eq({
duration_s: 0,
failures: []
})
end
end
context 'when top-level function raises error' do
it 'returns empty result with generic failure' do
allow(Gitlab::Prometheus::Internal).to receive(:prometheus_enabled?).and_raise(RuntimeError)
expect(subject[:topology]).to eq({
duration_s: 0,
failures: [
{ 'other' => 'RuntimeError' }
]
})
end
end
end
def receive_app_request_volume_query(result: nil)
receive(:query)
.with(/gitlab_usage_ping:ops:rate/)
.and_return(result || [
{
'metric' => { 'component' => 'http_requests', 'service' => 'workhorse' },
'value' => [1000, '0.01']
}
])
end
def receive_node_memory_query(result: nil)
receive(:query)
.with(/node_memory_total_bytes/, an_instance_of(Hash))
.and_return(result || [
{
'metric' => { 'instance' => 'instance1:8080' },
'value' => [1000, '512']
},
{
'metric' => { 'instance' => 'instance2:8090' },
'value' => [1000, '1024']
}
])
end
def receive_node_cpu_count_query(result: nil)
receive(:query)
.with(/node_cpus/, an_instance_of(Hash))
.and_return(result || [
{
'metric' => { 'instance' => 'instance2:8090' },
'value' => [1000, '16']
},
{
'metric' => { 'instance' => 'instance1:8080' },
'value' => [1000, '8']
}
])
end
def receive_node_uname_info_query(result: nil)
receive(:query)
.with('node_uname_info')
.and_return(result || [
{
"metric" => {
"__name__" => "node_uname_info",
"domainname" => "(none)",
"instance" => "instance1:9100",
"job" => "node_exporter",
"machine" => "x86_64",
"nodename" => "instance1",
"release" => "4.19.76-linuxkit",
"sysname" => "Linux"
},
"value" => [1592463033.359, "1"]
},
{
"metric" => {
"__name__" => "node_uname_info",
"domainname" => "(none)",
"instance" => "instance2:9100",
"job" => "node_exporter",
"machine" => "x86_64",
"nodename" => "instance2",
"release" => "4.15.0-101-generic",
"sysname" => "Linux"
},
"value" => [1592463033.359, "1"]
}
])
end
def receive_node_service_memory_rss_query(result: nil)
receive(:query)
.with(/process_resident_memory_bytes/, an_instance_of(Hash))
.and_return(result || [
{
'metric' => { 'instance' => 'instance1:8080', 'job' => 'gitlab-rails' },
'value' => [1000, '300']
},
{
'metric' => { 'instance' => 'instance1:8090', 'job' => 'gitlab-sidekiq' },
'value' => [1000, '303']
},
# instance 2: runs a dedicated Sidekiq + Redis (which uses a different metric name)
{
'metric' => { 'instance' => 'instance2:8090', 'job' => 'gitlab-sidekiq' },
'value' => [1000, '400']
},
{
'metric' => { 'instance' => 'instance2:9121', 'job' => 'redis' },
'value' => [1000, '402']
}
])
end
def receive_node_service_memory_uss_query(result: nil)
receive(:query)
.with(/process_unique_memory_bytes/, an_instance_of(Hash))
.and_return(result || [
{
'metric' => { 'instance' => 'instance1:8080', 'job' => 'gitlab-rails' },
'value' => [1000, '301']
}
])
end
def receive_node_service_memory_pss_query(result: nil)
receive(:query)
.with(/process_proportional_memory_bytes/, an_instance_of(Hash))
.and_return(result || [
{
'metric' => { 'instance' => 'instance1:8080', 'job' => 'gitlab-rails' },
'value' => [1000, '302']
},
{
'metric' => { 'instance' => 'instance2:8090', 'job' => 'gitlab-sidekiq' },
'value' => [1000, '401']
}
])
end
def receive_node_service_process_count_query(result: nil)
receive(:query)
.with(/service_process:count/, an_instance_of(Hash))
.and_return(result || [
# instance 1
{
'metric' => { 'instance' => 'instance1:8080', 'job' => 'gitlab-rails' },
'value' => [1000, '10']
},
{
'metric' => { 'instance' => 'instance1:8090', 'job' => 'gitlab-sidekiq' },
'value' => [1000, '5']
},
# instance 2
{
'metric' => { 'instance' => 'instance2:8090', 'job' => 'gitlab-sidekiq' },
'value' => [1000, '15']
},
{
'metric' => { 'instance' => 'instance2:9121', 'job' => 'redis' },
'value' => [1000, '1']
},
{
'metric' => { 'instance' => 'instance2:8080', 'job' => 'registry' },
'value' => [1000, '1']
},
# unknown service => should be stripped out
{
'metric' => { 'instance' => 'instance2:9000', 'job' => 'not-a-gitlab-service' },
'value' => [1000, '42']
}
])
end
def receive_node_service_app_server_workers_query(result: nil)
receive(:query)
.with(/app_server_workers/, an_instance_of(Hash))
.and_return(result || [
# instance 1
{
'metric' => { 'instance' => 'instance1:8080', 'job' => 'gitlab-rails', 'server' => 'puma' },
'value' => [1000, '2']
},
# instance 2
{
'metric' => { 'instance' => 'instance2:8080', 'job' => 'gitlab-rails', 'server' => 'unicorn' },
'value' => [1000, '1']
}
])
end
end