infrastructure/ansible
1
0
Fork 0
Code Pull requests 7 Activity Actions

copr-fe: prometheus monitoring to gather the actual state

Browse source 
Not the "historical" info which is calculated into the percentage
of time service spent in UP state.
This commit is contained in:
Pavel Raiskup 2023-02-10 14:40:39 +01:00
parent 39070dc595
commit cacaec9c8e
1 changed files with 27 additions and 37 deletions
Download patch file Download diff file Expand all files Collapse all files

64
roles/copr/frontend/files/monitoring.py
View file

@ -5,45 +5,35 @@ from bs4 import BeautifulSoup
from prometheus_client import CollectorRegistry, write_to_textfile, Gauge
class TextfileCollector:
NAGIOS_URL = 'https://nagios.fedoraproject.org/nagios/cgi-bin//avail.cgi'
def collect_nagios_service_state(url, name, documentation, filename):
registry = CollectorRegistry()
gauge = Gauge(name, documentation, registry=registry)
state = 0
def __init__(self, url, name, documentation, css_selector, filename):
self.registry = CollectorRegistry()
self.url = url
self.css_selector = css_selector
self.filename = filename
self.copr_cdn_status = Gauge(name, documentation, registry=self.registry)
try:
response = requests.get(url)
soup = BeautifulSoup(response.content, features="lxml")
if soup.select_one("div.serviceOK"):
state = 1
except Exception:
pass
def collect_data(self):
url = self.NAGIOS_URL + self.url
r = requests.get(url)
html = r.text
soup = BeautifulSoup(html, features="lxml")
service_ok = soup.select_one(self.css_selector)
return service_ok.text.strip('%')
def set(self, data):
self.copr_cdn_status.set(data)
write_to_textfile(self.filename, self.registry)
def collect(url, name, documentation, css_selector, filename):
textfile_collector = TextfileCollector(url, name, documentation, css_selector, filename)
percentage = textfile_collector.collect_data()
textfile_collector.set(percentage)
gauge.set(state)
write_to_textfile(filename, registry)
if __name__ == '__main__':
copr_cdn_url = "?show_log_entries=&full_log_entries=&" \
"host=copr-fe.aws.fedoraproject.org&service=The+copr+cdn+status&assumeinitialstates=yes&" \
"assumestateretention=yes&assumestatesduringnotrunning=yes&includesoftstates=no&" \
"initialassumedhoststate=0&initialassumedservicestate=0&timeperiod=last7days&backtrack=4"
copr_ping_url = "?show_log_entries=&host=copr-be.aws.fedoraproject.org&" \
"service=The+copr-ping+package+builds&assumeinitialstates=yes&assumestateretention=yes&" \
"assumestatesduringnotrunning=yes&includesoftstates=no&initialassumedhoststate=0&" \
"initialassumedservicestate=0&timeperiod=last7days&backtrack=4"
collect(copr_cdn_url, "copr_cdn_status", "Copr's CDN status", "td.serviceOK:nth-child(4)",
"/var/lib/prometheus/node-exporter/copr_cdn_status.prom.new")
collect(copr_ping_url, "copr_ping_status", "Status of build of copr-ping package", "td.serviceOK:nth-child(4)",
"/var/lib/prometheus/node-exporter/copr_ping_status.prom.new")
collect_nagios_service_state(
"https://nagios.fedoraproject.org/nagios/cgi-bin/extinfo.cgi?type=2&host=copr-fe.aws.fedoraproject.org&service=The+copr+cdn+status",
"fedora_copr_cdn_up",
"1 if Nagios reports that AWS CloudFront CDN for Fedora Copr works fine",
"/var/lib/prometheus/node-exporter/fedora_copr_cdn_up.prom",
)
collect_nagios_service_state(
"https://nagios.fedoraproject.org/nagios/cgi-bin/extinfo.cgi?type=2&host=copr-be.aws.fedoraproject.org&service=The+copr-ping+package+builds",
"fedora_copr_ping_package_ok",
"1 if Nagios reports that the \"ping\" package builds fine in Fedora Copr",
"/var/lib/prometheus/node-exporter/fedora_copr_ping_package_ok.prom",
)