fleet/server/datastore/mysql/cron_stats_test.go
Scott Gress 6bd9cc8a44
Monitor and alert on errors in cron jobs (#24347)
for #19930 

# Checklist for submitter

- [X] Changes file added for user-visible changes in `changes/`,
`orbit/changes/` or `ee/fleetd-chrome/changes`.
- [X] Input data is properly validated, `SELECT *` is avoided, SQL
injection is prevented (using placeholders for values in statements)
- [X] Added/updated tests
- [X] If database migrations are included, checked table schema to
confirm autoupdate
- [X] Manual QA for all new/changed functionality

# Details

This PR adds a new feature to the existing monitoring add-on. The add-on
will now send an SNS alert whenever a scheduled job like
"vulnerabilities" or "apple_mdm_apns_pusher" exits early due to errors.
The alert contains the job type and the set of errors (there can be
multiple, since jobs can have multiple sub-jobs). By default the SNS
topic for this new alert is the same as the one for the existing cron
system alerts, but it can be configured to use a separate topic (e.g.
dogfood instance will post to a separate slack channel).

The actual changes are:

**On the server side:**

- Add errors field to cron_stats table (json DEFAULT NULL)
- Added errors var to `Schedule` struct to collect errors from jobs
- In `RunAllJobs`, collect err from job into new errors var
- Update `Schedule.updateStats`and `CronStats.UpdateCronStats`to accept
errors argument
- If provided, update errors field of cron_stats table

**On the monitor side:**

- Add new SQL query to look for all completed schedules since last run
with non-null errors
- send SNS with job ID, name, errors

# Testing

New automated testing was added for the functional code that gathers and
stores errors from cron runs in the database. To test the actual Lambda,
I added a row in my `cron_stats` table with errors, then compiled and
ran the Lambda executable locally, pointing it to my local mysql and
localstack instances:

```
2024/12/03 14:43:54 main.go:258: Lambda execution environment not found.  Falling back to local execution.
2024/12/03 14:43:54 main.go:133: Connected to database!
2024/12/03 14:43:54 main.go:161: Row vulnerabilities last updated at 2024-11-27 03:30:03 +0000 UTC
2024/12/03 14:43:54 main.go:163: *** 1h hasn't updated in more than vulnerabilities, alerting! (status completed)
2024/12/03 14:43:54 main.go:70: Sending SNS Message
2024/12/03 14:43:54 main.go:74: Sending 'Environment: dev
Message: Fleet cron 'vulnerabilities' hasn't updated in more than 1h. Last status was 'completed' at 2024-11-27 03:30:03 +0000 UTC.' to 'arn:aws:sns:us-east-1:000000000000:topic1'
2024/12/03 14:43:54 main.go:82: {
  MessageId: "260864ff-4cc9-4951-acea-cef883b2de5f"
}
2024/12/03 14:43:54 main.go:198: *** mdm_apple_profile_manager job had errors, alerting! (errors {"something": "wrong"})
2024/12/03 14:43:54 main.go:70: Sending SNS Message
2024/12/03 14:43:54 main.go:74: Sending 'Environment: dev
Message: Fleet cron 'mdm_apple_profile_manager' (last updated 2024-12-03 20:34:14 +0000 UTC) raised errors during its run:
{"something": "wrong"}.' to 'arn:aws:sns:us-east-1:000000000000:topic1'
2024/12/03 14:43:54 main.go:82: {
  MessageId: "5cd085ef-89f6-42c1-8470-d80a22b295f8"
2024-12-19 15:55:29 -06:00

311 lines
11 KiB
Go

package mysql
import (
"context"
"database/sql"
"encoding/json"
"errors"
"testing"
"time"
"github.com/fleetdm/fleet/v4/server/fleet"
"github.com/jmoiron/sqlx"
"github.com/stretchr/testify/require"
)
type testCronStats struct {
fleet.CronStats
// Errors is a JSON string containing any errors encountered during the run.
Errors sql.NullString `db:"errors"`
}
func TestInsertUpdateCronStats(t *testing.T) {
const (
scheduleName = "test_sched"
instanceID = "test_instance"
)
ctx := context.Background()
ds := CreateMySQLDS(t)
id, err := ds.InsertCronStats(ctx, fleet.CronStatsTypeScheduled, scheduleName, instanceID, fleet.CronStatsStatusPending)
require.NoError(t, err)
res, err := ds.GetLatestCronStats(ctx, scheduleName)
require.NoError(t, err)
require.Len(t, res, 1)
require.Equal(t, id, res[0].ID)
require.Equal(t, fleet.CronStatsTypeScheduled, res[0].StatsType)
require.Equal(t, fleet.CronStatsStatusPending, res[0].Status)
err = ds.UpdateCronStats(ctx, id, fleet.CronStatsStatusCompleted, &fleet.CronScheduleErrors{
"some_job": errors.New("some error"),
"some_other_job": errors.New("some other error"),
})
require.NoError(t, err)
res, err = ds.GetLatestCronStats(ctx, scheduleName)
require.NoError(t, err)
require.Len(t, res, 1)
require.Equal(t, id, res[0].ID)
require.Equal(t, fleet.CronStatsTypeScheduled, res[0].StatsType)
require.Equal(t, fleet.CronStatsStatusCompleted, res[0].Status)
var stats []testCronStats
err = sqlx.SelectContext(ctx, ds.reader(ctx), &stats, `SELECT * FROM cron_stats ORDER BY id`)
require.NoError(t, err)
// Make sure we got valid JSON back.
var actualMap map[string]string
err = json.Unmarshal([]byte(stats[0].Errors.String), &actualMap)
require.NoError(t, err)
// Compare the error JSON with the expected object.
expectedJSON := `{"some_job": "some error", "some_other_job": "some other error"}`
var expectedMap map[string]string
err = json.Unmarshal([]byte(expectedJSON), &expectedMap)
require.NoError(t, err)
require.Equal(t, actualMap, expectedMap)
}
func TestGetLatestCronStats(t *testing.T) {
const (
scheduleName = "test_sched"
instanceID = "test_instance"
)
ctx := context.Background()
ds := CreateMySQLDS(t)
insertTestCS := func(name string, statsType fleet.CronStatsType, status fleet.CronStatsStatus, createdAt time.Time) {
stmt := `INSERT INTO cron_stats (stats_type, name, instance, status, created_at) VALUES (?, ?, ?, ?, ?)`
_, err := ds.writer(ctx).ExecContext(ctx, stmt, statsType, name, instanceID, status, createdAt)
require.NoError(t, err)
}
then := time.Now().UTC().Truncate(time.Second).Add(-24 * time.Hour)
// insert two "scheduled" stats
insertTestCS(scheduleName, fleet.CronStatsTypeScheduled, fleet.CronStatsStatusPending, then.Add(2*time.Minute))
insertTestCS(scheduleName, fleet.CronStatsTypeScheduled, fleet.CronStatsStatusCompleted, then.Add(1*time.Minute))
// most recent record is returned for "scheduled" stats type
res, err := ds.GetLatestCronStats(ctx, scheduleName)
require.NoError(t, err)
require.Len(t, res, 1)
require.Equal(t, fleet.CronStatsTypeScheduled, res[0].StatsType)
require.Equal(t, fleet.CronStatsStatusPending, res[0].Status)
require.Equal(t, then.Add(2*time.Minute), res[0].CreatedAt)
// insert two "triggered" stats
insertTestCS(scheduleName, fleet.CronStatsTypeTriggered, fleet.CronStatsStatusCompleted, then.Add(2*time.Hour))
insertTestCS(scheduleName, fleet.CronStatsTypeTriggered, fleet.CronStatsStatusCompleted, then.Add(1*time.Hour))
// most recent record is returned for both "scheduled" stats type and "triggered" stats type
res, err = ds.GetLatestCronStats(ctx, scheduleName)
require.NoError(t, err)
require.Len(t, res, 2)
require.Equal(t, fleet.CronStatsTypeScheduled, res[0].StatsType)
require.Equal(t, fleet.CronStatsStatusPending, res[0].Status)
require.Equal(t, then.Add(2*time.Minute), res[0].CreatedAt)
require.Equal(t, fleet.CronStatsTypeTriggered, res[1].StatsType)
require.Equal(t, fleet.CronStatsStatusCompleted, res[1].Status)
require.Equal(t, then.Add(2*time.Hour), res[1].CreatedAt)
// insert some other stats that shouldn't be returned
insertTestCS(scheduleName, fleet.CronStatsTypeScheduled, fleet.CronStatsStatusExpired, then.Add(3*time.Hour)) // expired status shouldn't be returned
insertTestCS(scheduleName, fleet.CronStatsTypeTriggered, fleet.CronStatsStatusExpired, then.Add(3*time.Hour)) // expired status shouldn't be returned
insertTestCS(scheduleName, fleet.CronStatsTypeScheduled, fleet.CronStatsStatusCanceled, then.Add(4*time.Hour)) // canceled status shouldn't be returned
insertTestCS(scheduleName, fleet.CronStatsTypeTriggered, fleet.CronStatsStatusCanceled, then.Add(4*time.Hour)) // canceled status shouldn't be returned
insertTestCS("schedule_1337", fleet.CronStatsTypeTriggered, fleet.CronStatsStatusPending, then.Add(5*time.Hour)) // different name shouldn't be returned
// most recent record is returned for both "scheduled" stats type and "triggered" stats type
res, err = ds.GetLatestCronStats(ctx, scheduleName)
require.NoError(t, err)
require.Len(t, res, 2)
require.Equal(t, fleet.CronStatsTypeScheduled, res[0].StatsType)
require.Equal(t, fleet.CronStatsStatusPending, res[0].Status)
require.Equal(t, then.Add(2*time.Minute), res[0].CreatedAt)
require.Equal(t, fleet.CronStatsTypeTriggered, res[1].StatsType)
require.Equal(t, fleet.CronStatsStatusCompleted, res[1].Status)
require.Equal(t, then.Add(2*time.Hour), res[1].CreatedAt)
}
func TestCleanupCronStats(t *testing.T) {
ctx := context.Background()
ds := CreateMySQLDS(t)
now := time.Now().UTC().Truncate(time.Second)
twoDaysAgo := now.Add(-2 * 24 * time.Hour)
name := "test_sched"
instance := "test_instance"
cases := []struct {
createdAt time.Time
status fleet.CronStatsStatus
shouldCleanupMaxPending bool
shouldCleanupMaxAge bool
}{
{
createdAt: now,
status: fleet.CronStatsStatusCompleted,
shouldCleanupMaxPending: false,
shouldCleanupMaxAge: false,
},
{
createdAt: now,
status: fleet.CronStatsStatusPending,
shouldCleanupMaxPending: false,
shouldCleanupMaxAge: false,
},
{
createdAt: now.Add(-1 * time.Hour),
status: fleet.CronStatsStatusPending,
shouldCleanupMaxPending: false,
shouldCleanupMaxAge: false,
},
{
createdAt: now.Add(-2 * time.Hour),
status: fleet.CronStatsStatusExpired,
shouldCleanupMaxPending: false,
shouldCleanupMaxAge: false,
},
{
createdAt: now.Add(-3 * time.Hour),
status: fleet.CronStatsStatusPending,
shouldCleanupMaxPending: true,
shouldCleanupMaxAge: false,
},
{
createdAt: now.Add(-3 * time.Hour),
status: fleet.CronStatsStatusCompleted,
shouldCleanupMaxPending: false,
shouldCleanupMaxAge: false,
},
{
createdAt: twoDaysAgo.Add(1 * time.Hour),
status: fleet.CronStatsStatusCompleted,
shouldCleanupMaxPending: false,
shouldCleanupMaxAge: false,
},
{
createdAt: twoDaysAgo.Add(-1 * time.Hour),
status: fleet.CronStatsStatusCompleted,
shouldCleanupMaxPending: false,
shouldCleanupMaxAge: true,
},
}
for _, c := range cases {
stmt := `INSERT INTO cron_stats (stats_type, name, instance, status, created_at) VALUES (?, ?, ?, ?, ?)`
_, err := ds.writer(ctx).ExecContext(ctx, stmt, fleet.CronStatsTypeScheduled, name, instance, c.status, c.createdAt)
require.NoError(t, err)
}
var stats []testCronStats
err := sqlx.SelectContext(ctx, ds.reader(ctx), &stats, `SELECT * FROM cron_stats ORDER BY id`)
require.NoError(t, err)
require.Len(t, stats, len(cases))
for i, s := range stats {
require.Equal(t, cases[i].createdAt, s.CreatedAt)
require.Equal(t, cases[i].status, s.Status)
}
err = ds.CleanupCronStats(ctx)
require.NoError(t, err)
stats = []testCronStats{}
err = sqlx.SelectContext(ctx, ds.reader(ctx), &stats, `SELECT * FROM cron_stats ORDER BY id`)
require.NoError(t, err)
require.Len(t, stats, len(cases)-1) // case[7] was deleted because it exceeded max age
for i, c := range cases {
if i >= len(stats) {
require.True(t, c.shouldCleanupMaxAge)
break
}
if c.shouldCleanupMaxPending {
require.Equal(t, fleet.CronStatsStatusExpired, stats[i].Status)
} else {
require.Equal(t, c.status, stats[i].Status)
}
}
}
func TestUpdateAllCronStatsForInstance(t *testing.T) {
ctx := context.Background()
ds := CreateMySQLDS(t)
cases := []struct {
instance string
schedName string
status fleet.CronStatsStatus
shouldUpdate bool
}{
{
instance: "inst1",
schedName: "sched1",
status: fleet.CronStatsStatusCompleted,
shouldUpdate: false,
},
{
instance: "inst1",
schedName: "sched1",
status: fleet.CronStatsStatusPending,
shouldUpdate: true,
},
{
instance: "inst1",
schedName: "sched2",
status: fleet.CronStatsStatusExpired,
shouldUpdate: false,
},
{
instance: "inst1",
schedName: "sched2",
status: fleet.CronStatsStatusPending,
shouldUpdate: true,
},
{
instance: "inst2",
schedName: "sched1",
status: fleet.CronStatsStatusPending,
shouldUpdate: false,
},
{
instance: "inst2",
schedName: "sched2",
status: fleet.CronStatsStatusPending,
shouldUpdate: false,
},
}
for _, c := range cases {
stmt := `INSERT INTO cron_stats (stats_type, name, instance, status) VALUES (?, ?, ?, ?)`
_, err := ds.writer(ctx).ExecContext(ctx, stmt, fleet.CronStatsTypeScheduled, c.schedName, c.instance, c.status)
require.NoError(t, err)
}
var stats []testCronStats
err := sqlx.SelectContext(ctx, ds.reader(ctx), &stats, `SELECT * FROM cron_stats ORDER BY id`)
require.NoError(t, err)
require.Len(t, stats, len(cases))
for i, s := range stats {
require.Equal(t, cases[i].schedName, s.Name)
require.Equal(t, cases[i].instance, s.Instance)
require.Equal(t, cases[i].status, s.Status)
}
err = ds.UpdateAllCronStatsForInstance(ctx, "inst1", fleet.CronStatsStatusPending, fleet.CronStatsStatusCanceled)
require.NoError(t, err)
stats = []testCronStats{}
err = sqlx.SelectContext(ctx, ds.reader(ctx), &stats, `SELECT * FROM cron_stats ORDER BY id`)
require.NoError(t, err)
require.Len(t, stats, len(cases))
for i, c := range cases {
s := stats[i]
require.Equal(t, c.instance, s.Instance)
require.Equal(t, c.schedName, s.Name)
if c.shouldUpdate {
require.Equal(t, fleet.CronStatsStatusCanceled, s.Status)
} else {
require.Equal(t, c.status, s.Status)
}
}
}