feat(storage): Support 30d badges (#836)

* feat(storage): Add support for 30d uptime badge

Fix #714

* Fix typo

* Fix test

* Fix typo

* Improve implementation

* Add check in existing test

* Add extra test to ensure functionality works

* Add support for 30d response time chart too
This commit is contained in:
TwiN 2024-08-11 22:40:19 -04:00 committed by GitHub
parent 90ffea9fb6
commit dd435a8eaf
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
8 changed files with 276 additions and 27 deletions

View File

@ -2127,7 +2127,7 @@ The path to generate a badge is the following:
/api/v1/endpoints/{key}/uptimes/{duration}/badge.svg /api/v1/endpoints/{key}/uptimes/{duration}/badge.svg
``` ```
Where: Where:
- `{duration}` is `7d`, `24h` or `1h` - `{duration}` is `30d` (alpha), `7d`, `24h` or `1h`
- `{key}` has the pattern `<GROUP_NAME>_<ENDPOINT_NAME>` in which both variables have ` `, `/`, `_`, `,` and `.` replaced by `-`. - `{key}` has the pattern `<GROUP_NAME>_<ENDPOINT_NAME>` in which both variables have ` `, `/`, `_`, `,` and `.` replaced by `-`.
For instance, if you want the uptime during the last 24 hours from the endpoint `frontend` in the group `core`, For instance, if you want the uptime during the last 24 hours from the endpoint `frontend` in the group `core`,
@ -2192,7 +2192,7 @@ The endpoint to generate a badge is the following:
/api/v1/endpoints/{key}/response-times/{duration}/badge.svg /api/v1/endpoints/{key}/response-times/{duration}/badge.svg
``` ```
Where: Where:
- `{duration}` is `7d`, `24h` or `1h` - `{duration}` is `30d` (alpha), `7d`, `24h` or `1h`
- `{key}` has the pattern `<GROUP_NAME>_<ENDPOINT_NAME>` in which both variables have ` `, `/`, `_`, `,` and `.` replaced by `-`. - `{key}` has the pattern `<GROUP_NAME>_<ENDPOINT_NAME>` in which both variables have ` `, `/`, `_`, `,` and `.` replaced by `-`.

View File

@ -37,11 +37,13 @@ var (
// UptimeBadge handles the automatic generation of badge based on the group name and endpoint name passed. // UptimeBadge handles the automatic generation of badge based on the group name and endpoint name passed.
// //
// Valid values for :duration -> 7d, 24h, 1h // Valid values for :duration -> 30d, 7d, 24h, 1h
func UptimeBadge(c *fiber.Ctx) error { func UptimeBadge(c *fiber.Ctx) error {
duration := c.Params("duration") duration := c.Params("duration")
var from time.Time var from time.Time
switch duration { switch duration {
case "30d":
from = time.Now().Add(-30 * 24 * time.Hour)
case "7d": case "7d":
from = time.Now().Add(-7 * 24 * time.Hour) from = time.Now().Add(-7 * 24 * time.Hour)
case "24h": case "24h":
@ -49,7 +51,7 @@ func UptimeBadge(c *fiber.Ctx) error {
case "1h": case "1h":
from = time.Now().Add(-2 * time.Hour) // Because uptime metrics are stored by hour, we have to cheat a little from = time.Now().Add(-2 * time.Hour) // Because uptime metrics are stored by hour, we have to cheat a little
default: default:
return c.Status(400).SendString("Durations supported: 7d, 24h, 1h") return c.Status(400).SendString("Durations supported: 30d, 7d, 24h, 1h")
} }
key := c.Params("key") key := c.Params("key")
uptime, err := store.Get().GetUptimeByKey(key, from, time.Now()) uptime, err := store.Get().GetUptimeByKey(key, from, time.Now())
@ -69,12 +71,14 @@ func UptimeBadge(c *fiber.Ctx) error {
// ResponseTimeBadge handles the automatic generation of badge based on the group name and endpoint name passed. // ResponseTimeBadge handles the automatic generation of badge based on the group name and endpoint name passed.
// //
// Valid values for :duration -> 7d, 24h, 1h // Valid values for :duration -> 30d, 7d, 24h, 1h
func ResponseTimeBadge(cfg *config.Config) fiber.Handler { func ResponseTimeBadge(cfg *config.Config) fiber.Handler {
return func(c *fiber.Ctx) error { return func(c *fiber.Ctx) error {
duration := c.Params("duration") duration := c.Params("duration")
var from time.Time var from time.Time
switch duration { switch duration {
case "30d":
from = time.Now().Add(-30 * 24 * time.Hour)
case "7d": case "7d":
from = time.Now().Add(-7 * 24 * time.Hour) from = time.Now().Add(-7 * 24 * time.Hour)
case "24h": case "24h":
@ -82,7 +86,7 @@ func ResponseTimeBadge(cfg *config.Config) fiber.Handler {
case "1h": case "1h":
from = time.Now().Add(-2 * time.Hour) // Because response time metrics are stored by hour, we have to cheat a little from = time.Now().Add(-2 * time.Hour) // Because response time metrics are stored by hour, we have to cheat a little
default: default:
return c.Status(400).SendString("Durations supported: 7d, 24h, 1h") return c.Status(400).SendString("Durations supported: 30d, 7d, 24h, 1h")
} }
key := c.Params("key") key := c.Params("key")
averageResponseTime, err := store.Get().GetAverageResponseTimeByKey(key, from, time.Now()) averageResponseTime, err := store.Get().GetAverageResponseTimeByKey(key, from, time.Now())
@ -161,6 +165,8 @@ func HealthBadgeShields(c *fiber.Ctx) error {
func generateUptimeBadgeSVG(duration string, uptime float64) []byte { func generateUptimeBadgeSVG(duration string, uptime float64) []byte {
var labelWidth, valueWidth, valueWidthAdjustment int var labelWidth, valueWidth, valueWidthAdjustment int
switch duration { switch duration {
case "30d":
labelWidth = 70
case "7d": case "7d":
labelWidth = 65 labelWidth = 65
case "24h": case "24h":
@ -227,6 +233,8 @@ func getBadgeColorFromUptime(uptime float64) string {
func generateResponseTimeBadgeSVG(duration string, averageResponseTime int, key string, cfg *config.Config) []byte { func generateResponseTimeBadgeSVG(duration string, averageResponseTime int, key string, cfg *config.Config) []byte {
var labelWidth, valueWidth int var labelWidth, valueWidth int
switch duration { switch duration {
case "30d":
labelWidth = 110
case "7d": case "7d":
labelWidth = 105 labelWidth = 105
case "24h": case "24h":

View File

@ -32,14 +32,18 @@ var (
func ResponseTimeChart(c *fiber.Ctx) error { func ResponseTimeChart(c *fiber.Ctx) error {
duration := c.Params("duration") duration := c.Params("duration")
chartTimestampFormatter := chart.TimeValueFormatterWithFormat(timeFormat)
var from time.Time var from time.Time
switch duration { switch duration {
case "30d":
from = time.Now().Truncate(time.Hour).Add(-30 * 24 * time.Hour)
chartTimestampFormatter = chart.TimeDateValueFormatter
case "7d": case "7d":
from = time.Now().Truncate(time.Hour).Add(-24 * 7 * time.Hour) from = time.Now().Truncate(time.Hour).Add(-7 * 24 * time.Hour)
case "24h": case "24h":
from = time.Now().Truncate(time.Hour).Add(-24 * time.Hour) from = time.Now().Truncate(time.Hour).Add(-24 * time.Hour)
default: default:
return c.Status(400).SendString("Durations supported: 7d, 24h") return c.Status(400).SendString("Durations supported: 30d, 7d, 24h")
} }
hourlyAverageResponseTime, err := store.Get().GetHourlyAverageResponseTimeByKey(c.Params("key"), from, time.Now()) hourlyAverageResponseTime, err := store.Get().GetHourlyAverageResponseTimeByKey(c.Params("key"), from, time.Now())
if err != nil { if err != nil {
@ -88,7 +92,7 @@ func ResponseTimeChart(c *fiber.Ctx) error {
Width: 1280, Width: 1280,
Height: 300, Height: 300,
XAxis: chart.XAxis{ XAxis: chart.XAxis{
ValueFormatter: chart.TimeValueFormatterWithFormat(timeFormat), ValueFormatter: chartTimestampFormatter,
GridMajorStyle: gridStyle, GridMajorStyle: gridStyle,
GridMinorStyle: gridStyle, GridMinorStyle: gridStyle,
Style: axisStyle, Style: axisStyle,

View File

@ -49,6 +49,11 @@ func TestResponseTimeChart(t *testing.T) {
Path: "/api/v1/endpoints/core_frontend/response-times/7d/chart.svg", Path: "/api/v1/endpoints/core_frontend/response-times/7d/chart.svg",
ExpectedCode: http.StatusOK, ExpectedCode: http.StatusOK,
}, },
{
Name: "chart-response-time-30d",
Path: "/api/v1/endpoints/core_frontend/response-times/30d/chart.svg",
ExpectedCode: http.StatusOK,
},
{ {
Name: "chart-response-time-with-invalid-duration", Name: "chart-response-time-with-invalid-duration",
Path: "/api/v1/endpoints/core_backend/response-times/3d/chart.svg", Path: "/api/v1/endpoints/core_backend/response-times/3d/chart.svg",

View File

@ -7,8 +7,8 @@ import (
) )
const ( const (
numberOfHoursInTenDays = 10 * 24 uptimeCleanUpThreshold = 32 * 24
sevenDays = 7 * 24 * time.Hour uptimeRetention = 30 * 24 * time.Hour
) )
// processUptimeAfterResult processes the result by extracting the relevant from the result and recalculating the uptime // processUptimeAfterResult processes the result by extracting the relevant from the result and recalculating the uptime
@ -30,10 +30,10 @@ func processUptimeAfterResult(uptime *endpoint.Uptime, result *endpoint.Result)
hourlyStats.TotalExecutionsResponseTime += uint64(result.Duration.Milliseconds()) hourlyStats.TotalExecutionsResponseTime += uint64(result.Duration.Milliseconds())
// Clean up only when we're starting to have too many useless keys // 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 // 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. // 32 days, despite the fact that we are deleting everything that's older than 30 days.
// This is to prevent re-iterating on every `processUptimeAfterResult` as soon as the uptime has been logged for 7 days. // This is to prevent re-iterating on every `processUptimeAfterResult` as soon as the uptime has been logged for 30 days.
if len(uptime.HourlyStatistics) > numberOfHoursInTenDays { if len(uptime.HourlyStatistics) > uptimeCleanUpThreshold {
sevenDaysAgo := time.Now().Add(-(sevenDays + time.Hour)).Unix() sevenDaysAgo := time.Now().Add(-(uptimeRetention + time.Hour)).Unix()
for hourlyUnixTimestamp := range uptime.HourlyStatistics { for hourlyUnixTimestamp := range uptime.HourlyStatistics {
if sevenDaysAgo > hourlyUnixTimestamp { if sevenDaysAgo > hourlyUnixTimestamp {
delete(uptime.HourlyStatistics, hourlyUnixTimestamp) delete(uptime.HourlyStatistics, hourlyUnixTimestamp)

View File

@ -51,8 +51,8 @@ func TestAddResultUptimeIsCleaningUpAfterItself(t *testing.T) {
timestamp := now.Add(-12 * 24 * time.Hour) timestamp := now.Add(-12 * 24 * time.Hour)
for timestamp.Unix() <= now.Unix() { for timestamp.Unix() <= now.Unix() {
AddResult(status, &endpoint.Result{Timestamp: timestamp, Success: true}) AddResult(status, &endpoint.Result{Timestamp: timestamp, Success: true})
if len(status.Uptime.HourlyStatistics) > numberOfHoursInTenDays { if len(status.Uptime.HourlyStatistics) > uptimeCleanUpThreshold {
t.Errorf("At no point in time should there be more than %d entries in status.SuccessfulExecutionsPerHour, but there are %d", numberOfHoursInTenDays, len(status.Uptime.HourlyStatistics)) t.Errorf("At no point in time should there be more than %d entries in status.SuccessfulExecutionsPerHour, but there are %d", uptimeCleanUpThreshold, len(status.Uptime.HourlyStatistics))
} }
// Simulate endpoint with an interval of 3 minutes // Simulate endpoint with an interval of 3 minutes
timestamp = timestamp.Add(3 * time.Minute) timestamp = timestamp.Add(3 * time.Minute)

View File

@ -28,11 +28,13 @@ const (
// for aesthetic purposes, I deemed it wasn't worth the performance impact of yet another one-to-many table. // for aesthetic purposes, I deemed it wasn't worth the performance impact of yet another one-to-many table.
arraySeparator = "|~|" arraySeparator = "|~|"
uptimeCleanUpThreshold = 10 * 24 * time.Hour // Maximum uptime age before triggering a cleanup
eventsCleanUpThreshold = common.MaximumNumberOfEvents + 10 // Maximum number of events before triggering a cleanup eventsCleanUpThreshold = common.MaximumNumberOfEvents + 10 // Maximum number of events before triggering a cleanup
resultsCleanUpThreshold = common.MaximumNumberOfResults + 10 // Maximum number of results before triggering a cleanup resultsCleanUpThreshold = common.MaximumNumberOfResults + 10 // Maximum number of results before triggering a cleanup
uptimeRetention = 7 * 24 * time.Hour uptimeTotalEntriesMergeThreshold = 100 // Maximum number of uptime entries before triggering a merge
uptimeAgeCleanUpThreshold = 32 * 24 * time.Hour // Maximum uptime age before triggering a cleanup
uptimeRetention = 30 * 24 * time.Hour // Minimum duration that must be kept to operate as intended
uptimeHourlyBuffer = 48 * time.Hour // Number of hours to buffer from now when determining which hourly uptime entries can be merged into daily uptime entries
cacheTTL = 10 * time.Minute cacheTTL = 10 * time.Minute
) )
@ -322,12 +324,27 @@ func (s *Store) Insert(ep *endpoint.Endpoint, result *endpoint.Result) error {
if err = s.updateEndpointUptime(tx, endpointID, result); err != nil { if err = s.updateEndpointUptime(tx, endpointID, result); err != nil {
log.Printf("[sql.Insert] Failed to update uptime for endpoint with key=%s: %s", ep.Key(), err.Error()) log.Printf("[sql.Insert] Failed to update uptime for endpoint with key=%s: %s", ep.Key(), err.Error())
} }
// Clean up old uptime entries // Merge hourly uptime entries that can be merged into daily entries and clean up old uptime entries
numberOfUptimeEntries, err := s.getNumberOfUptimeEntriesByEndpointID(tx, endpointID)
if err != nil {
log.Printf("[sql.Insert] Failed to retrieve total number of uptime entries for endpoint with key=%s: %s", ep.Key(), err.Error())
} else {
// Merge older hourly uptime entries into daily uptime entries if we have more than uptimeTotalEntriesMergeThreshold
if numberOfUptimeEntries >= uptimeTotalEntriesMergeThreshold {
log.Printf("[sql.Insert] Merging hourly uptime entries for endpoint with key=%s; This is a lot of work, it shouldn't happen too often", ep.Key())
if err = s.mergeHourlyUptimeEntriesOlderThanMergeThresholdIntoDailyUptimeEntries(tx, endpointID); err != nil {
log.Printf("[sql.Insert] Failed to merge hourly uptime entries for endpoint with key=%s: %s", ep.Key(), err.Error())
}
}
}
// Clean up outdated uptime entries
// In most cases, this would be handled by mergeHourlyUptimeEntriesOlderThanMergeThresholdIntoDailyUptimeEntries,
// but if Gatus was temporarily shut down, we might have some old entries that need to be cleaned up
ageOfOldestUptimeEntry, err := s.getAgeOfOldestEndpointUptimeEntry(tx, endpointID) ageOfOldestUptimeEntry, err := s.getAgeOfOldestEndpointUptimeEntry(tx, endpointID)
if err != nil { if err != nil {
log.Printf("[sql.Insert] Failed to retrieve oldest endpoint uptime entry for endpoint with key=%s: %s", ep.Key(), err.Error()) log.Printf("[sql.Insert] Failed to retrieve oldest endpoint uptime entry for endpoint with key=%s: %s", ep.Key(), err.Error())
} else { } else {
if ageOfOldestUptimeEntry > uptimeCleanUpThreshold { if ageOfOldestUptimeEntry > uptimeAgeCleanUpThreshold {
if err = s.deleteOldUptimeEntries(tx, endpointID, time.Now().Add(-(uptimeRetention + time.Hour))); err != nil { if err = s.deleteOldUptimeEntries(tx, endpointID, time.Now().Add(-(uptimeRetention + time.Hour))); err != nil {
log.Printf("[sql.Insert] Failed to delete old uptime entries for endpoint with key=%s: %s", ep.Key(), err.Error()) log.Printf("[sql.Insert] Failed to delete old uptime entries for endpoint with key=%s: %s", ep.Key(), err.Error())
} }
@ -865,6 +882,12 @@ func (s *Store) getNumberOfResultsByEndpointID(tx *sql.Tx, endpointID int64) (in
return numberOfResults, err return numberOfResults, err
} }
func (s *Store) getNumberOfUptimeEntriesByEndpointID(tx *sql.Tx, endpointID int64) (int64, error) {
var numberOfUptimeEntries int64
err := tx.QueryRow("SELECT COUNT(1) FROM endpoint_uptimes WHERE endpoint_id = $1", endpointID).Scan(&numberOfUptimeEntries)
return numberOfUptimeEntries, err
}
func (s *Store) getAgeOfOldestEndpointUptimeEntry(tx *sql.Tx, endpointID int64) (time.Duration, error) { func (s *Store) getAgeOfOldestEndpointUptimeEntry(tx *sql.Tx, endpointID int64) (time.Duration, error) {
rows, err := tx.Query( rows, err := tx.Query(
` `
@ -948,6 +971,92 @@ func (s *Store) deleteOldUptimeEntries(tx *sql.Tx, endpointID int64, maxAge time
return err return err
} }
// mergeHourlyUptimeEntriesOlderThanMergeThresholdIntoDailyUptimeEntries merges all hourly uptime entries older than
// uptimeHourlyMergeThreshold from now into daily uptime entries by summing all hourly entries of the same day into a
// single entry.
//
// This effectively limits the number of uptime entries to (48+(n-2)) where 48 is for the first 48 entries with hourly
// entries (defined by uptimeHourlyBuffer) and n is the number of days for all entries older than 48 hours.
// Supporting 30d of entries would then result in far less than 24*30=720 entries.
func (s *Store) mergeHourlyUptimeEntriesOlderThanMergeThresholdIntoDailyUptimeEntries(tx *sql.Tx, endpointID int64) error {
// Calculate timestamp of the first full day of uptime entries that would not impact the uptime calculation for 24h badges
// The logic is that once at least 48 hours passed, we:
// - No longer need to worry about keeping hourly entries
// - Don't have to worry about new hourly entries being inserted, as the day has already passed
// which implies that no matter at what hour of the day we are, any timestamp + 48h floored to the current day
// will never impact the 24h uptime badge calculation
now := time.Now()
minThreshold := now.Add(-uptimeHourlyBuffer)
minThreshold = time.Date(minThreshold.Year(), minThreshold.Month(), minThreshold.Day(), 0, 0, 0, 0, minThreshold.Location())
maxThreshold := now.Add(-uptimeRetention)
// Get all uptime entries older than uptimeHourlyMergeThreshold
rows, err := tx.Query(
`
SELECT hour_unix_timestamp, total_executions, successful_executions, total_response_time
FROM endpoint_uptimes
WHERE endpoint_id = $1
AND hour_unix_timestamp < $2
AND hour_unix_timestamp >= $3
`,
endpointID,
minThreshold.Unix(),
maxThreshold.Unix(),
)
if err != nil {
return err
}
type Entry struct {
totalExecutions int
successfulExecutions int
totalResponseTime int
}
dailyEntries := make(map[int64]*Entry)
for rows.Next() {
var unixTimestamp int64
entry := Entry{}
if err = rows.Scan(&unixTimestamp, &entry.totalExecutions, &entry.successfulExecutions, &entry.totalResponseTime); err != nil {
return err
}
timestamp := time.Unix(unixTimestamp, 0)
unixTimestampFlooredAtDay := time.Date(timestamp.Year(), timestamp.Month(), timestamp.Day(), 0, 0, 0, 0, timestamp.Location()).Unix()
if dailyEntry := dailyEntries[unixTimestampFlooredAtDay]; dailyEntry == nil {
dailyEntries[unixTimestampFlooredAtDay] = &entry
} else {
dailyEntries[unixTimestampFlooredAtDay].totalExecutions += entry.totalExecutions
dailyEntries[unixTimestampFlooredAtDay].successfulExecutions += entry.successfulExecutions
dailyEntries[unixTimestampFlooredAtDay].totalResponseTime += entry.totalResponseTime
}
}
// Delete older hourly uptime entries
_, err = tx.Exec("DELETE FROM endpoint_uptimes WHERE endpoint_id = $1 AND hour_unix_timestamp < $2", endpointID, minThreshold.Unix())
if err != nil {
return err
}
// Insert new daily uptime entries
for unixTimestamp, entry := range dailyEntries {
_, err = tx.Exec(
`
INSERT INTO endpoint_uptimes (endpoint_id, hour_unix_timestamp, total_executions, successful_executions, total_response_time)
VALUES ($1, $2, $3, $4, $5)
ON CONFLICT(endpoint_id, hour_unix_timestamp) DO UPDATE SET
total_executions = $3,
successful_executions = $4,
total_response_time = $5
`,
endpointID,
unixTimestamp,
entry.totalExecutions,
entry.successfulExecutions,
entry.totalResponseTime,
)
if err != nil {
return err
}
}
// TODO: Find a way to ignore entries that were already merged?
return nil
}
func generateCacheKey(endpointKey string, p *paging.EndpointStatusParams) string { func generateCacheKey(endpointKey string, p *paging.EndpointStatusParams) string {
return fmt.Sprintf("%s-%d-%d-%d-%d", endpointKey, p.EventsPage, p.EventsPageSize, p.ResultsPage, p.ResultsPageSize) return fmt.Sprintf("%s-%d-%d-%d-%d", endpointKey, p.EventsPage, p.EventsPageSize, p.ResultsPage, p.ResultsPageSize)
} }

View File

@ -2,6 +2,7 @@ package sql
import ( import (
"errors" "errors"
"fmt"
"testing" "testing"
"time" "time"
@ -132,18 +133,18 @@ func TestStore_InsertCleansUpOldUptimeEntriesProperly(t *testing.T) {
} }
// Since this is one hour before reaching the clean up threshold, the oldest entry should now be this one // Since this is one hour before reaching the clean up threshold, the oldest entry should now be this one
store.Insert(&testEndpoint, &endpoint.Result{Timestamp: now.Add(-(uptimeCleanUpThreshold - time.Hour)), Success: true}) store.Insert(&testEndpoint, &endpoint.Result{Timestamp: now.Add(-(uptimeAgeCleanUpThreshold - time.Hour)), Success: true})
tx, _ = store.db.Begin() tx, _ = store.db.Begin()
oldest, _ = store.getAgeOfOldestEndpointUptimeEntry(tx, 1) oldest, _ = store.getAgeOfOldestEndpointUptimeEntry(tx, 1)
_ = tx.Commit() _ = tx.Commit()
if oldest.Truncate(time.Hour) != uptimeCleanUpThreshold-time.Hour { if oldest.Truncate(time.Hour) != uptimeAgeCleanUpThreshold-time.Hour {
t.Errorf("oldest endpoint uptime entry should've been ~%s hours old, was %s", uptimeCleanUpThreshold-time.Hour, oldest) t.Errorf("oldest endpoint uptime entry should've been ~%s hours old, was %s", uptimeAgeCleanUpThreshold-time.Hour, oldest)
} }
// Since this entry is after the uptimeCleanUpThreshold, both this entry as well as the previous // Since this entry is after the uptimeAgeCleanUpThreshold, both this entry as well as the previous
// one should be deleted since they both surpass uptimeRetention // one should be deleted since they both surpass uptimeRetention
store.Insert(&testEndpoint, &endpoint.Result{Timestamp: now.Add(-(uptimeCleanUpThreshold + time.Hour)), Success: true}) store.Insert(&testEndpoint, &endpoint.Result{Timestamp: now.Add(-(uptimeAgeCleanUpThreshold + time.Hour)), Success: true})
tx, _ = store.db.Begin() tx, _ = store.db.Begin()
oldest, _ = store.getAgeOfOldestEndpointUptimeEntry(tx, 1) oldest, _ = store.getAgeOfOldestEndpointUptimeEntry(tx, 1)
@ -153,8 +154,128 @@ func TestStore_InsertCleansUpOldUptimeEntriesProperly(t *testing.T) {
} }
} }
func TestStore_HourlyUptimeEntriesAreMergedIntoDailyUptimeEntriesProperly(t *testing.T) {
store, _ := NewStore("sqlite", t.TempDir()+"/TestStore_HourlyUptimeEntriesAreMergedIntoDailyUptimeEntriesProperly.db", false)
defer store.Close()
now := time.Now().Truncate(time.Hour)
now = time.Date(now.Year(), now.Month(), now.Day(), now.Hour(), 0, 0, 0, now.Location())
scenarios := []struct {
numberOfHours int
expectedMaxUptimeEntries int64
}{
{numberOfHours: 1, expectedMaxUptimeEntries: 1},
{numberOfHours: 10, expectedMaxUptimeEntries: 10},
{numberOfHours: 50, expectedMaxUptimeEntries: 50},
{numberOfHours: 75, expectedMaxUptimeEntries: 75},
{numberOfHours: 99, expectedMaxUptimeEntries: 99},
{numberOfHours: 150, expectedMaxUptimeEntries: 100},
{numberOfHours: 300, expectedMaxUptimeEntries: 100},
{numberOfHours: 768, expectedMaxUptimeEntries: 100}, // 32 days (in hours), which means anything beyond that won't be persisted anyway
{numberOfHours: 1000, expectedMaxUptimeEntries: 100},
}
// Note that is not technically an accurate real world representation, because uptime entries are always added in
// the present, while this test is inserting results from the past to simulate long term uptime entries.
// Since we want to test the behavior and not the test itself, this is a "best effort" approach.
for _, scenario := range scenarios {
t.Run(fmt.Sprintf("num-hours-%d-expected-max-entries-%d", scenario.numberOfHours, scenario.expectedMaxUptimeEntries), func(t *testing.T) {
for i := scenario.numberOfHours; i > 0; i-- {
//fmt.Printf("i: %d (%s)\n", i, now.Add(-time.Duration(i)*time.Hour))
// Create an uptime entry
err := store.Insert(&testEndpoint, &endpoint.Result{Timestamp: now.Add(-time.Duration(i) * time.Hour), Success: true})
if err != nil {
t.Log(err)
}
//// DEBUGGING: check number of uptime entries for endpoint
//tx, _ := store.db.Begin()
//numberOfUptimeEntriesForEndpoint, err := store.getNumberOfUptimeEntriesByEndpointID(tx, 1)
//if err != nil {
// t.Log(err)
//}
//_ = tx.Commit()
//t.Logf("i=%d; numberOfHours=%d; There are currently %d uptime entries for endpointID=%d", i, scenario.numberOfHours, numberOfUptimeEntriesForEndpoint, 1)
}
// check number of uptime entries for endpoint
tx, _ := store.db.Begin()
numberOfUptimeEntriesForEndpoint, err := store.getNumberOfUptimeEntriesByEndpointID(tx, 1)
if err != nil {
t.Log(err)
}
_ = tx.Commit()
//t.Logf("numberOfHours=%d; There are currently %d uptime entries for endpointID=%d", scenario.numberOfHours, numberOfUptimeEntriesForEndpoint, 1)
if scenario.expectedMaxUptimeEntries < numberOfUptimeEntriesForEndpoint {
t.Errorf("expected %d (uptime entries) to be smaller than %d", numberOfUptimeEntriesForEndpoint, scenario.expectedMaxUptimeEntries)
}
store.Clear()
})
}
}
func TestStore_getEndpointUptime(t *testing.T) {
store, _ := NewStore("sqlite", t.TempDir()+"/TestStore_InsertCleansUpEventsAndResultsProperly.db", false)
defer store.Clear()
defer store.Close()
// Add 768 hourly entries (32 days)
// Daily entries should be merged from hourly entries automatically
for i := 768; i > 0; i-- {
err := store.Insert(&testEndpoint, &endpoint.Result{Timestamp: time.Now().Add(-time.Duration(i) * time.Hour), Duration: time.Second, Success: true})
if err != nil {
t.Log(err)
}
}
// Check the number of uptime entries
tx, _ := store.db.Begin()
numberOfUptimeEntriesForEndpoint, err := store.getNumberOfUptimeEntriesByEndpointID(tx, 1)
if err != nil {
t.Log(err)
}
if numberOfUptimeEntriesForEndpoint < 20 || numberOfUptimeEntriesForEndpoint > 200 {
t.Errorf("expected number of uptime entries to be between 20 and 200, got %d", numberOfUptimeEntriesForEndpoint)
}
// Retrieve uptime for the past 30d
uptime, avgResponseTime, err := store.getEndpointUptime(tx, 1, time.Now().Add(-(30 * 24 * time.Hour)), time.Now())
if err != nil {
t.Log(err)
}
_ = tx.Commit()
if avgResponseTime != time.Second {
t.Errorf("expected average response time to be %s, got %s", time.Second, avgResponseTime)
}
if uptime != 1 {
t.Errorf("expected uptime to be 1, got %f", uptime)
}
// Add a new unsuccessful result, which should impact the uptime
err = store.Insert(&testEndpoint, &endpoint.Result{Timestamp: time.Now(), Duration: time.Second, Success: false})
if err != nil {
t.Log(err)
}
// Retrieve uptime for the past 30d
tx, _ = store.db.Begin()
uptime, _, err = store.getEndpointUptime(tx, 1, time.Now().Add(-(30 * 24 * time.Hour)), time.Now())
if err != nil {
t.Log(err)
}
_ = tx.Commit()
if uptime == 1 {
t.Errorf("expected uptime to be less than 1, got %f", uptime)
}
// Retrieve uptime for the past 30d, but excluding the last 24h
// This is not a real use case as there is no way for users to exclude the last 24h, but this is a great way
// to ensure that hourly merging works as intended
tx, _ = store.db.Begin()
uptimeExcludingLast24h, _, err := store.getEndpointUptime(tx, 1, time.Now().Add(-(30 * 24 * time.Hour)), time.Now().Add(-24*time.Hour))
if err != nil {
t.Log(err)
}
_ = tx.Commit()
if uptimeExcludingLast24h == uptime {
t.Error("expected uptimeExcludingLast24h to to be different from uptime, got")
}
}
func TestStore_InsertCleansUpEventsAndResultsProperly(t *testing.T) { func TestStore_InsertCleansUpEventsAndResultsProperly(t *testing.T) {
store, _ := NewStore("sqlite", t.TempDir()+"/TestStore_InsertCleansUpEventsAndResultsProperly.db", false) store, _ := NewStore("sqlite", t.TempDir()+"/TestStore_InsertCleansUpEventsAndResultsProperly.db", false)
defer store.Clear()
defer store.Close() defer store.Close()
for i := 0; i < resultsCleanUpThreshold+eventsCleanUpThreshold; i++ { for i := 0; i < resultsCleanUpThreshold+eventsCleanUpThreshold; i++ {
store.Insert(&testEndpoint, &testSuccessfulResult) store.Insert(&testEndpoint, &testSuccessfulResult)
@ -167,7 +288,6 @@ func TestStore_InsertCleansUpEventsAndResultsProperly(t *testing.T) {
t.Errorf("number of events shouldn't have exceeded %d, reached %d", eventsCleanUpThreshold, len(ss.Events)) t.Errorf("number of events shouldn't have exceeded %d, reached %d", eventsCleanUpThreshold, len(ss.Events))
} }
} }
store.Clear()
} }
func TestStore_InsertWithCaching(t *testing.T) { func TestStore_InsertWithCaching(t *testing.T) {
@ -218,6 +338,9 @@ func TestStore_Persistence(t *testing.T) {
if uptime, _ := store.GetUptimeByKey(testEndpoint.Key(), time.Now().Add(-time.Hour*24*7), time.Now()); uptime != 0.5 { if uptime, _ := store.GetUptimeByKey(testEndpoint.Key(), time.Now().Add(-time.Hour*24*7), time.Now()); uptime != 0.5 {
t.Errorf("the uptime over the past 7d should've been 0.5, got %f", uptime) t.Errorf("the uptime over the past 7d should've been 0.5, got %f", uptime)
} }
if uptime, _ := store.GetUptimeByKey(testEndpoint.Key(), time.Now().Add(-time.Hour*24*30), time.Now()); uptime != 0.5 {
t.Errorf("the uptime over the past 30d should've been 0.5, got %f", uptime)
}
ssFromOldStore, _ := store.GetEndpointStatus(testEndpoint.Group, testEndpoint.Name, paging.NewEndpointStatusParams().WithResults(1, common.MaximumNumberOfResults).WithEvents(1, common.MaximumNumberOfEvents)) ssFromOldStore, _ := store.GetEndpointStatus(testEndpoint.Group, testEndpoint.Name, paging.NewEndpointStatusParams().WithResults(1, common.MaximumNumberOfResults).WithEvents(1, common.MaximumNumberOfEvents))
if ssFromOldStore == nil || ssFromOldStore.Group != "group" || ssFromOldStore.Name != "name" || len(ssFromOldStore.Events) != 3 || len(ssFromOldStore.Results) != 2 { if ssFromOldStore == nil || ssFromOldStore.Group != "group" || ssFromOldStore.Name != "name" || len(ssFromOldStore.Events) != 3 || len(ssFromOldStore.Results) != 2 {
store.Close() store.Close()