extern/gatus
1
0
Fork 0
mirror of https://github.com/TwiN/gatus.git synced 2024-11-22 07:53:38 +01:00
Code Issues Packages Projects Releases Wiki Activity

Significantly improve uptime calculation

Browse Source
This commit is contained in:
TwinProduction 2021-03-04 23:00:30 -05:00
parent 8fe9d013b5
commit 6320237326
3 changed files with 19 additions and 33 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
core/service-status_test.go
View File

@ -26,7 +26,7 @@ func TestServiceStatus_AddResult(t *testing.T) {
serviceStatus.AddResult(&Result{Timestamp: time.Now()})
}
if len(serviceStatus.Results) != MaximumNumberOfResults {
t.Errorf("expected serviceStatus.Results to not exceed a length of 20")
t.Errorf("expected serviceStatus.Results to not exceed a length of %d", MaximumNumberOfResults)
}
}

47
core/uptime.go
View File

@ -1,15 +1,10 @@
package core
import (
"log"
"time"
)
const (
// RFC3339WithoutMinutesAndSeconds is the format defined by RFC3339 (see time.RFC3339) but with the minutes
// and seconds hardcoded to 0.
RFC3339WithoutMinutesAndSeconds = "2006-01-02T15:00:00Z07:00"
numberOfHoursInTenDays = 10 * 24
sevenDays = 7 * 24 * time.Hour
)
@ -25,47 +20,39 @@ type Uptime struct {
// LastHour is the uptime percentage over the past hour
LastHour float64 `json:"1h"`
// SuccessCountPerHour is a map containing the number of successes per hour, per timestamp following the
// custom RFC3339WithoutMinutesAndSeconds format
SuccessCountPerHour map[string]uint64 `json:"-"`
// SuccessCountPerHour is a map containing the number of successes (value) for every hourly unix timestamps (key)
SuccessCountPerHour map[int64]uint64 `json:"-"`
// TotalCountPerHour is a map containing the total number of checks per hour, per timestamp following the
// custom RFC3339WithoutMinutesAndSeconds format
TotalCountPerHour map[string]uint64 `json:"-"`
// TotalCountPerHour is a map containing the total number of checks (value) for every hourly unix timestamps (key)
TotalCountPerHour map[int64]uint64 `json:"-"`
}
// NewUptime creates a new Uptime
func NewUptime() *Uptime {
return &Uptime{
SuccessCountPerHour: make(map[string]uint64),
TotalCountPerHour: make(map[string]uint64),
SuccessCountPerHour: make(map[int64]uint64),
TotalCountPerHour: make(map[int64]uint64),
}
}
// ProcessResult processes the result by extracting the relevant from the result and recalculating the uptime
// if necessary
func (uptime *Uptime) ProcessResult(result *Result) {
timestampDateWithHour := result.Timestamp.Format(RFC3339WithoutMinutesAndSeconds)
unixTimestampFlooredAtHour := result.Timestamp.Unix() - (result.Timestamp.Unix() % 3600)
if result.Success {
uptime.SuccessCountPerHour[timestampDateWithHour]++
uptime.SuccessCountPerHour[unixTimestampFlooredAtHour]++
}
uptime.TotalCountPerHour[timestampDateWithHour]++
uptime.TotalCountPerHour[unixTimestampFlooredAtHour]++
// Clean up only when we're starting to have too many useless keys
// Note that this is only triggered when there are more entries than there should be after
// 10 days, despite the fact that we are deleting everything that's older than 7 days.
// This is to prevent re-iterating on every `ProcessResult` as soon as the uptime has been logged for 7 days.
if len(uptime.TotalCountPerHour) > numberOfHoursInTenDays {
sevenDaysAgo := time.Now().Add(-(sevenDays + time.Hour))
for k := range uptime.TotalCountPerHour {
dateWithHour, err := time.Parse(time.RFC3339, k)
if err != nil {
// This shouldn't happen, but we'll log it in case it does happen
log.Println("[uptime][ProcessResult] Failed to parse programmatically generated timestamp:", err.Error())
continue
}
if sevenDaysAgo.Unix() > dateWithHour.Unix() {
delete(uptime.TotalCountPerHour, k)
delete(uptime.SuccessCountPerHour, k)
sevenDaysAgo := time.Now().Add(-(sevenDays + time.Hour)).Unix()
for hourlyUnixTimestamp := range uptime.TotalCountPerHour {
if sevenDaysAgo > hourlyUnixTimestamp {
delete(uptime.TotalCountPerHour, hourlyUnixTimestamp)
delete(uptime.SuccessCountPerHour, hourlyUnixTimestamp)
}
}
}
@ -92,9 +79,9 @@ func (uptime *Uptime) recalculate() {
// The oldest uptime bracket starts 7 days ago, so we'll start from there
timestamp := now.Add(-sevenDays)
for now.Sub(timestamp) >= 0 {
timestampDateWithHour := timestamp.Format(RFC3339WithoutMinutesAndSeconds)
successCountForTimestamp := uptime.SuccessCountPerHour[timestampDateWithHour]
totalCountForTimestamp := uptime.TotalCountPerHour[timestampDateWithHour]
hourlyUnixTimestamp := timestamp.Unix() - (timestamp.Unix() % 3600)
successCountForTimestamp := uptime.SuccessCountPerHour[hourlyUnixTimestamp]
totalCountForTimestamp := uptime.TotalCountPerHour[hourlyUnixTimestamp]
uptimeBrackets["7d_success"] += successCountForTimestamp
uptimeBrackets["7d_total"] += totalCountForTimestamp
if now.Sub(timestamp) <= 24*time.Hour {

3
core/uptime_test.go
View File

@ -52,9 +52,8 @@ func TestServiceStatus_AddResultUptimeIsCleaningUpAfterItself(t *testing.T) {
for timestamp.Unix() <= now.Unix() {
serviceStatus.AddResult(&Result{Timestamp: timestamp, Success: true})
if len(serviceStatus.Uptime.SuccessCountPerHour) > numberOfHoursInTenDays {
t.Errorf("At no point in time should there be more than %d entries in serviceStatus.SuccessCountPerHour", numberOfHoursInTenDays)
t.Errorf("At no point in time should there be more than %d entries in serviceStatus.SuccessCountPerHour, but there are %d", numberOfHoursInTenDays, len(serviceStatus.Uptime.SuccessCountPerHour))
}
//fmt.Printf("timestamp=%s; uptimeDuringLastHour=%f; timeAgo=%s\n", timestamp.Format(time.RFC3339), serviceStatus.UptimeDuringLastHour, time.Since(timestamp))
if now.Sub(timestamp) > time.Hour && serviceStatus.Uptime.LastHour != 0 {
t.Error("most recent timestamp > 1h ago, expected serviceStatus.Uptime.LastHour to be 0, got", serviceStatus.Uptime.LastHour)
}