- moved files from root to VictoriaMetrics folder to be able to mount operator docs and VictoriaMetrics docs independently - added ability to run website locally The following checks are **mandatory**: - [ ] My change adheres [VictoriaMetrics contributing guidelines](https://docs.victoriametrics.com/contributing/).
17 KiB
sort | weight | title | menu | aliases | |||||||
---|---|---|---|---|---|---|---|---|---|---|---|
200 | 200 | Scrape config examples |
|
|
Static configs
Let's start from a simple case with scraping targets at pre-defined addresses.
Create a scrape.yaml
file with the following contents:
scrape_configs:
- job_name: node-exporter
static_configs:
- targets:
- localhost:9100
After you created the scrape.yaml
file, download and unpack single-node VictoriaMetrics to the same directory:
wget https://github.com/VictoriaMetrics/VictoriaMetrics/releases/download/v1.102.0/victoria-metrics-linux-amd64-v1.102.0.tar.gz
tar xzf victoria-metrics-linux-amd64-v1.102.0.tar.gz
Then start VictoriaMetrics and instruct it to scrape targets defined in scrape.yaml
and save scraped metrics
to local storage according to these docs:
./victoria-metrics-prod -promscrape.config=scrape.yaml
Now open the http://localhost:8428/targets
page in web browser in order to see the current status for scrape targets.
The page must contain the information about the target at http://localhost:9100/metrics
url.
It is likely the target has state: down
if you didn't start node-exporter
on localhost
.
Let's add a new scrape config to scrape.yaml
for scraping VictoriaMetrics metrics:
scrape_configs:
- job_name: node-exporter
static_configs:
- targets:
- localhost:9100
- job_name: victoriametrics
static_configs:
- targets:
- http://localhost:8428/metrics
Note that the last specified target contains the full url instead of host and port.
This is an extension supported by VictoriaMetrics and vmagent - you can use both host:port
and full urls in scrape target lists.
Send SIGHUP
signal victoria-metrics-prod
process, so it reloads the updated scrape.yaml
:
kill -HUP `pidof victoria-metrics-prod`
Now the http://localhost:8428/targets
page must contain two targets - http://localhost:9100/metrics
and http://localhost:8428/metrics
.
The last one should have state: up
, since this is VictoriaMetrics itself.
Let's query the scraped metrics. Open http://localhost:8428/vmui/
aka vmui, enter up
in the query input field
and press enter
. You'll see a graph for up
metrics. It must contain two lines for the targets defined in scrape.yaml
file above.
See these docs about up
metric. You can explore other scraped metrics
in vmui
via Prometheus metrics explorer.
Let's look closely to the contents of the scrape.yaml
file created above:
scrape_configs:
- job_name: node-exporter
static_configs:
- targets:
- localhost:9100
- job_name: victoriametrics
static_configs:
- targets:
- http://localhost:8428/metrics
The scrape_configs
section contains a list of scrape configs.
Our scrape.yaml
file contains two scrape configs - for job_name: node-exporter
and for job_name: victoriametrics
.
vmagent and single-node VictoriaMetrics
can efficiently process thousands of scrape configs in production.
Every scrape config in the list must contain job_name
field - its' value is used as job
label
in all the metrics scraped from targets defined in this scrape config.
Every scrape config must contain at least a single section from this list.
Every scrape config may contain other options described here.
In our case only static_configs
sections are used.
These sections consist of a list of static configs according to these docs.
Every static config contains a list of targets
, which need to be scraped. The target address is used as instance
label in all the metrics scraped from the target.
vmagent and single-node VictoriaMetrics can efficiently process tens of thousands of targets in production. If you need scraping more targets, then see these docs.
Targets are scraped at http
or https
urls, which are formed according to these rules.
It is possible to modify scrape urls via relabeling if needed.
File-based target discovery
It may be not so convenient updating scrape.yaml
file with static_configs
every time new scrape target is added, changed or removed. In this case file_sd_configs
can come to rescue. It allows defining a list of scrape targets in JSON
files, and automatically updating the list of scrape targets
at vmagent or single-node VictoriaMetrics side
when the corresponding JSON
files are updated.
Let's create node_exporter_targets.json
file with the following contents:
[
{
"targets": ["host1:9100", "host2:9100"]
}
]
Then create scrape.yaml
file with the following contents:
scrape_configs:
- job_name: node-exporter
file_sd_configs:
- files:
- node_exporter_targets.json
Then start single-node VictoriaMetrics according to these docs:
# Download and unpack single-node VictoriaMetrics
wget https://github.com/VictoriaMetrics/VictoriaMetrics/releases/download/v1.102.0/victoria-metrics-linux-amd64-v1.102.0.tar.gz
tar xzf victoria-metrics-linux-amd64-v1.102.0.tar.gz
# Run single-node VictoriaMetrics with the given scrape.yaml
./victoria-metrics-prod -promscrape.config=scrape.yaml
Then open http://localhost:8428/targets
page in web browser and see that it contains the two targets defined in node_exporter_targets.json
above.
Now let's add more targets to node_exporter_targets.json
:
[
{
"targets": ["host1:9100", "host2:9100", "http://host3:9100/metrics", "http://host4:9100/metrics"]
}
]
Note that the added targets contains full urls instead of host and port.
This is an extension supported by VictoriaMetrics and vmagent - you can use both host:port
and full urls in scrape target lists.
Save the updated node_exporter_targets.json
, wait for 30 seconds and then refresh the http://localhost:8428/targets
page.
Now this page must contain all the targets defined in the updated node_exporter_targets.json
.
By default vmagent and single-node VictoriaMetrics
check for updates in files
specified at file_sd_configs
every 30 seconds. This interval can be changed via -promscrape.fileSDCheckInterval
command-line flag.
For example, the following command starts VictoriaMetrics, which checks for updates in file_sd_configs
every 5 seconds:
./victoria-metrics-prod -promscrape.config=scrape.yaml -promscrape.fileSDCheckInterval=5s
If the files
contents is broken during the check, then the previous list of scrape targets is kept.
It is possible specifying http
and/or https
urls in files
list. For example, the following config instructs
obtaining fresh list of targets at http://central-config-server/targets?type=node-exporter
url
additionally to node_exporter_targets.json
local file:
scrape_configs:
- job_name: node-exporter
file_sd_configs:
- files:
- node_exporter_targets.json
- 'http://central-config-server/targets?type=node-exporter'
It is possible to specify directories with *
wildcards for distinct sets of targets at file_sd_configs
.
See these docs for details.
vmagent and single-node VictoriaMetrics can efficiently scrape tens of thousands of scrape targets. If you need scraping more targets, then see these docs.
Targets are scraped at http
or https
urls, which are formed according to these rules.
It is possible to modify scrape urls via relabeling if needed.
HTTP-based target discovery
It may not so convenient maintaining a list of local files for file_sd_configs
.
In this case http_sd_configs
can help.
They allow specifying a list of http
or https
urls, which return targets, which need to be scraped.
For example, the following -promscrape.config
periodically fetches the list of targets from the specified url:
scrape_configs:
- job_name: node-exporter
http_sd_configs:
- url: "http://central-config-server/targets?type=node-exporter"
Kubernetes target discovery
Kubernetes target discovery is non-trivial task in general. That's why it is recommended using either victoria-metrics-k8s-stack Helm chart or VictoriaMetrics operator for Kubernetes for Kubernetes monitoring.
If you feel brave, let's look at a few typical cases for Kubernetes monitoring.
Discovering and scraping node-exporter
targets in Kubernetes
The following -promscrape.config
instructs discovering and scraping all the node-exporter
targets inside Kubernetes cluster:
scrape_configs:
- job_name: node-exporter
kubernetes_sd_configs:
- role: pod
relabel_configs:
# Leave only targets with `node-exporter` container name.
# If node-exporter containers have another name in your Kubernetes cluster,
# then adjust the regex value accordingly.
#
- source_labels: [__meta_kubernetes_pod_container_name]
regex: node-exporter
action: keep
# Copy node name into `node` label, so node-exporter targets
# can be attributed to a particular node.
#
- source_labels: [__meta_kubernetes_pod_node_name]
target_label: node
See kubernetes_sd_configs
docs for more details.
See relabeling docs for details on relabel_configs
.
Discovering and scraping kube-state-metrics
in Kubernetes
kube-state-metrics is a special metrics exporter,
which exposes state
metrics for all the Kubernetes objects such as container
, pod
, node
, etc.
It already sets namespace
, container
, pod
and node
labels for every exposed metric,
so these metrics shouldn't be set in target relabeling.
The following -promscrape.config
instructs discovering and scraping kube-state-metrics target inside Kubernetes cluster:
scrape_configs:
- job_name: kube-state-metrics
kubernetes_sd_configs:
- role: pod
relabel_configs:
# Leave only targets with `kube-state-metrics` container name.
# If kube-state-metrics container has another name in your Kubernetes cluster,
# then adjust the regex value accordingly.
- source_labels: [__meta_kubernetes_pod_container_name]
regex: kube-state-metrics
action: keep
# kube-state-metrics container may expose multiple ports.
# We need scraping only the e.g. service port, and do not need scraping e.g. telemetry port.
# The kube-state-metrics service port usually equals to 8080.
# Modify the regex accordingly if you use other port for kube-state-metrics.
#
- source_labels: [__meta_kubernetes_pod_container_port_number]
regex: "8080"
action: keep
See kubernetes_sd_configs
docs for more details.
See relabeling docs for details on relabel_configs
.
Discovering and scraping metrics from cadvisor
cadvisor exposes resource usage metrics for every container in Kubernetes.
The following -promscrape.config
can be used for collecting cadvisor
metrics in Kubernetes:
scrape_configs:
- job_name: cadvisor
kubernetes_sd_configs:
# Cadvisor is installed on every Kubernetes node, so use `role: node` service discovery
#
- role: node
# This is needed for scraping cadvisor metrics from Kubernetes API server proxy.
# See relabel_configs below.
#
bearer_token_file: /var/run/secrets/kubernetes.io/serviceaccount/token
tls_config:
ca_file: /var/run/secrets/kubernetes.io/serviceaccount/ca.crt
relabel_configs:
# Cadvisor metrics are better to scrape from Kubernetes API server proxy.
# There is no need to add container, pod and node labels to the scraped metrics,
# since cadvisor adds these labels on itself.
#
- source_labels: [__meta_kubernetes_node_name]
target_label: __address__
regex: '(.+)'
replacement: https://kubernetes.default.svc/api/v1/nodes/$1/proxy/metrics/cadvisor
- source_labels: [__meta_kubernetes_node_name]
target_label: instance
See kubernetes_sd_configs
docs for more details.
See relabeling docs for details on relabel_configs
.
See these docs for details on bearer_token_file
and tls_config
options.
Discovering and scraping metrics for a particular container in Kubernetes
The following -promscrape.config
instructs discovering and scraping metrics for all the containers with the name my-super-app
.
It is expected that these containers expose only a single TCP port, which serves its metrics at /metrics
page
according to Prometheus text exposition format:
scrape_configs:
- job_name: my-super-app
kubernetes_sd_configs:
- role: pod
relabel_configs:
# Leave only targets with the container name, which matches the `job_name` specified above
# See https://docs.victoriametrics.com/relabeling/#how-to-modify-instance-and-job for details on `job` label.
#
- source_labels: [__meta_kubernetes_pod_container_name]
target_label: job
action: keepequal
# Keep namespace, node, pod and container labels, so they can be used
# for joining additional `state` labels exposed by kube-state-metrics
# for the particular target.
#
- source_labels: [__meta_kubernetes_namespace]
target_label: namespace
- source_labels: [__meta_kubernetes_pod_node_name]
target_label: node
- source_labels: [__meta_kubernetes_pod_name]
target_label: pod
- source_labels: [__meta_kubernetes_pod_container_name]
target_label: container
See kubernetes_sd_configs
docs for more details.
See relabeling docs for details on relabel_configs
.