postfinance / discovery Goto Github PK

The following happens if a config file is not readable:

discovery --help
panic: ~/.config/discovery/config.yaml: open /home/user/.config/discovery/config.yaml: permission denied

goroutine 1 [running]:
github.com/alecthomas/kong.Parse(0xacf6c0, 0xc000260000, 0xc000184b00, 0x8, 0x8, 0x22)
        /home/rz/golang/pkg/mod/github.com/alecthomas/[email protected]/global.go:11 +0x145
main.main()

Created upstream pull request: alecthomas/kong#163

In order that systemd post script is always successfull

Otherwise prometheus is not happy:

Mar 15 16:37:53 example.com  prometheus[54858]: level=error ts=2021-03-15T15:37:53.780Z caller=file.go:237 component="discovery manager scrape" discovery=file msg="Error adding file watch" path=/appl/prometheus/service-discovery/p1-example.com/systemd err="no such file or directory"

The option "--selector-filter" does not work as the match is executed against labels and not against the selector field of the service:

// ServiceBySelector filters Services by Selector.
func ServiceBySelector(selector labels.Selector) FilterFunc {
	return func(s Service) bool {
                return selector.Matches(s.Labels)
	}
}

<configured-directory>/servers/<servername>/<namespace>/default/<jobname>.json
<configured-directory>/servers/<servername>/<namespace>/blackbox/<jobname>.json

Für prometheus mit namen p1-prom-alsu001.pnet.ch gibt es dann folgende Konfig:

scrape_configs:
  - job_name: 'default'
    scrape_interval: 15s
    file_sd_configs:
      - files: ['/appl/prometheus/etc/service-discovery/p1-prom-alsu001.pnet.ch/default/*']
  - job_name: 'k8s_kubelet'
    scrape_interval: 15s
    file_sd_configs:
      - files: ['/appl/prometheus/etc/service-discovery/p1-prom-alsu001.pnet.ch/blackbox/*']

Es gibt zwei Registrierungsarten:

• default
• blackbox

• --name, -N: filter by job name
• --selector. -s: filter by label selector
• --endpoint, -e: filter by endpoint
• --server, -S: filter by server

Implement the http_sd

This is showed:

discovery svc register -h
Usage: discovery service register --oidc-endpoint=STRING --oidc-client-id=STRING --endpoint=STRING --namespace="default" [<name>]

instead of:

discovery svc register -h
Usage: discovery service register --oidc-endpoint=STRING --oidc-client-id=STRING --endpoint=STRING --namespace="default" <name>

This should be fixed upstream: alecthomas/kong#153

example situation: 2 prometheus instances promA and promB, 4 endpoints. Endpoints A, B and C have 400 samples, while endpoint D has 3'000.

Currently, endpoints will be distributed to something like this:

promA: A + B = 800 samples
promB: C + D = 3'400 samples

ideally, endpoints will be distributed to better distribute samples instead of endpoints...

promA: A + B + C = 1'200 samples
promB: D = 3'000 samples

...or a reasonable mix thereof

• ratelimiter on write
• wenn limit erreicht in writeChan channel schreiben und write ignorieren
• atomic pending write setzen
• lesen von pendingWrite und rate.Wait ausführen -> wenn erreicht: write
• atomic pending write entfernen

When a service is unregistered, the exporter logs the following error:

2021-05-18T10:43:37.913+0200        ERROR        exporter/exporter.go:188        failed to delete service        {"namespace": "systemd", "id": "190196c8-5ba9-5a4c-a6db-8a017d88180b", "err": "not found"

for services which the exporter is not configured for. This error occurs, when the exporter wants to delete a service form its cache. But since the exporter only caches services it one prometheus server, this service never existed in this cache.

Trim leading (and trailing?) spaces from labels while registering (or correctly tokenize input).

E.g. following register command has a label ("label2") which has a leading space after registering with prometheus:

discovery service register -e http://example.com/metrics example --labels=label1=value1, label2=value2

Scraping in Prometheus won't work and logs the following error:

Apr 20 11:11:40 foobar prometheus[536382]: level=error ts=2021-04-20T09:11:40.078Z caller=file.go:344 component="discovery manager scrape" discovery=file msg="Error reading file" path=/appl/prometheus/service-discovery/foobar/appl/MY_APP.json err="\" label2\" is not a valid label name"

Maybe, the same issue also applies to label values.