fleet/server/service/live_queries.go
Victor Lyuboslavsky 1a679d0882
Moving context.WithoutCancel outside defer. (#17260)
#17197

Fixing orphaned live queries when context is canceled

Co-authored-by: Lucas Rodriguez <lucas@fleetdm.com>
2024-02-29 13:39:27 -03:00

413 lines
13 KiB
Go

package service
import (
"context"
"errors"
"fmt"
"os"
"strconv"
"strings"
"sync"
"time"
"github.com/fleetdm/fleet/v4/server"
"github.com/fleetdm/fleet/v4/server/authz"
"github.com/fleetdm/fleet/v4/server/contexts/ctxerr"
"github.com/fleetdm/fleet/v4/server/contexts/logging"
"github.com/fleetdm/fleet/v4/server/fleet"
"github.com/fleetdm/fleet/v4/server/ptr"
"github.com/go-kit/log/level"
)
type runLiveQueryRequest struct {
QueryIDs []uint `json:"query_ids"`
HostIDs []uint `json:"host_ids"`
}
type runOneLiveQueryRequest struct {
QueryID uint `url:"id"`
HostIDs []uint `json:"host_ids"`
}
type runLiveQueryOnHostRequest struct {
Identifier string `url:"identifier"`
Query string `json:"query"`
}
type runLiveQueryOnHostByIDRequest struct {
HostID uint `url:"id"`
Query string `json:"query"`
}
type summaryPayload struct {
TargetedHostCount int `json:"targeted_host_count"`
RespondedHostCount int `json:"responded_host_count"`
}
type runLiveQueryResponse struct {
Summary summaryPayload `json:"summary"`
Err error `json:"error,omitempty"`
Results []fleet.QueryCampaignResult `json:"live_query_results"`
}
func (r runLiveQueryResponse) error() error { return r.Err }
type runOneLiveQueryResponse struct {
QueryID uint `json:"query_id"`
TargetedHostCount int `json:"targeted_host_count"`
RespondedHostCount int `json:"responded_host_count"`
Results []fleet.QueryResult `json:"results"`
Err error `json:"error,omitempty"`
}
func (r runOneLiveQueryResponse) error() error { return r.Err }
type runLiveQueryOnHostResponse struct {
HostID uint `json:"host_id"`
Rows []map[string]string `json:"rows"`
Query string `json:"query"`
Status fleet.HostStatus `json:"status"`
Error string `json:"error,omitempty"`
}
func (r runLiveQueryOnHostResponse) error() error { return nil }
func runOneLiveQueryEndpoint(ctx context.Context, request interface{}, svc fleet.Service) (errorer, error) {
req := request.(*runOneLiveQueryRequest)
// Only allow a host to be specified once in HostIDs
hostIDs := server.RemoveDuplicatesFromSlice(req.HostIDs)
campaignResults, respondedHostCount, err := runLiveQuery(ctx, svc, []uint{req.QueryID}, "", hostIDs)
if err != nil {
return nil, err
}
//goland:noinspection GoPreferNilSlice -- use an empty slice here so that API returns an empty array if there are no results
queryResults := []fleet.QueryResult{}
if len(campaignResults) > 0 {
if campaignResults[0].Err != nil {
return nil, campaignResults[0].Err
}
if campaignResults[0].Results != nil {
queryResults = campaignResults[0].Results
}
}
res := runOneLiveQueryResponse{
QueryID: req.QueryID,
TargetedHostCount: len(hostIDs),
RespondedHostCount: respondedHostCount,
Results: queryResults,
}
return res, nil
}
func runLiveQueryEndpoint(ctx context.Context, request interface{}, svc fleet.Service) (errorer, error) {
req := request.(*runLiveQueryRequest)
// Only allow a query to be specified once
queryIDs := server.RemoveDuplicatesFromSlice(req.QueryIDs)
// Only allow a host to be specified once in HostIDs
hostIDs := server.RemoveDuplicatesFromSlice(req.HostIDs)
queryResults, respondedHostCount, err := runLiveQuery(ctx, svc, queryIDs, "", hostIDs)
if err != nil {
return nil, err
}
res := runLiveQueryResponse{
Summary: summaryPayload{
TargetedHostCount: len(hostIDs),
RespondedHostCount: respondedHostCount,
},
Results: queryResults,
}
return res, nil
}
func runLiveQueryOnHostEndpoint(ctx context.Context, request interface{}, svc fleet.Service) (errorer, error) {
req := request.(*runLiveQueryOnHostRequest)
host, err := svc.HostLiteByIdentifier(ctx, req.Identifier)
if err != nil {
return nil, ctxerr.Wrap(ctx, badRequest(fmt.Sprintf("host not found: %s: %s", req.Identifier, err.Error())))
}
return runLiveQueryOnHost(svc, ctx, host, req.Query)
}
func runLiveQueryOnHostByIDEndpoint(ctx context.Context, request interface{}, svc fleet.Service) (errorer, error) {
req := request.(*runLiveQueryOnHostByIDRequest)
host, err := svc.HostLiteByID(ctx, req.HostID)
if err != nil {
return nil, ctxerr.Wrap(ctx, badRequest(fmt.Sprintf("host not found: %d: %s", req.HostID, err.Error())))
}
return runLiveQueryOnHost(svc, ctx, host, req.Query)
}
func runLiveQueryOnHost(svc fleet.Service, ctx context.Context, host *fleet.HostLite, query string) (errorer, error) {
query = strings.TrimSpace(query)
if query == "" {
return nil, ctxerr.Wrap(ctx, badRequest("query is required"))
}
res := runLiveQueryOnHostResponse{
HostID: host.ID,
Query: query,
}
status := (&fleet.Host{
DistributedInterval: host.DistributedInterval,
ConfigTLSRefresh: host.ConfigTLSRefresh,
SeenTime: host.SeenTime,
}).Status(time.Now())
switch status {
case fleet.StatusOnline, fleet.StatusNew:
res.Status = fleet.StatusOnline
case fleet.StatusOffline, fleet.StatusMIA, fleet.StatusMissing:
res.Status = fleet.StatusOffline
return res, nil
default:
return nil, fmt.Errorf("unknown host status: %s", status)
}
queryResults, _, err := runLiveQuery(ctx, svc, []uint{0}, query, []uint{host.ID})
if err != nil {
return nil, err
}
if len(queryResults) > 0 {
var err error
if queryResults[0].Err != nil {
err = queryResults[0].Err
} else if len(queryResults[0].Results) > 0 {
queryResult := queryResults[0].Results[0]
if queryResult.Error != nil {
err = errors.New(*queryResult.Error)
}
res.Rows = queryResult.Rows
res.HostID = queryResult.HostID
} else {
err = errors.New("timeout waiting for results")
}
if err != nil {
res.Error = err.Error()
}
}
return res, nil
}
func runLiveQuery(ctx context.Context, svc fleet.Service, queryIDs []uint, query string, hostIDs []uint) (
[]fleet.QueryCampaignResult, int, error,
) {
// The period used here should always be less than the request timeout for any load
// balancer/proxy between Fleet and the API client.
period := os.Getenv("FLEET_LIVE_QUERY_REST_PERIOD")
if period == "" {
period = "25s"
}
duration, err := time.ParseDuration(period)
if err != nil {
duration = 25 * time.Second
logging.WithExtras(ctx, "live_query_rest_period_err", err)
}
queryResults, respondedHostCount, err := svc.RunLiveQueryDeadline(ctx, queryIDs, query, hostIDs, duration)
if err != nil {
return nil, 0, err
}
// Check if all query results were forbidden due to lack of authorization.
allResultsForbidden := len(queryResults) > 0 && respondedHostCount == 0
if allResultsForbidden {
for _, r := range queryResults {
if r.Error == nil || *r.Error != authz.ForbiddenErrorMessage {
allResultsForbidden = false
break
}
}
}
if allResultsForbidden {
return nil, 0, authz.ForbiddenWithInternal(
"All Live Query results were forbidden.", authz.UserFromContext(ctx), nil, nil,
)
}
return queryResults, respondedHostCount, nil
}
func (svc *Service) RunLiveQueryDeadline(
ctx context.Context, queryIDs []uint, query string, hostIDs []uint, deadline time.Duration,
) ([]fleet.QueryCampaignResult, int, error) {
if len(queryIDs) == 0 || len(hostIDs) == 0 {
svc.authz.SkipAuthorization(ctx)
return nil, 0, ctxerr.Wrap(ctx, badRequest("query_ids and host_ids are required"))
}
wg := sync.WaitGroup{}
resultsCh := make(chan fleet.QueryCampaignResult)
counterMutex := sync.Mutex{}
respondedHostIDs := make(map[uint]struct{})
for _, queryID := range queryIDs {
queryID := queryID
wg.Add(1)
go func() {
defer wg.Done()
queryIDPtr := &queryID
queryString := ""
// 0 is a special ID that indicates we should use raw SQL query instead
if queryID == 0 {
queryIDPtr = nil
queryString = query
}
campaign, err := svc.NewDistributedQueryCampaign(ctx, queryString, queryIDPtr, fleet.HostTargets{HostIDs: hostIDs})
if err != nil {
level.Error(svc.logger).Log(
"msg", "new distributed query campaign",
"queryString", queryString,
"queryID", queryID,
"err", err,
)
resultsCh <- fleet.QueryCampaignResult{QueryID: queryID, Error: ptr.String(err.Error()), Err: err}
return
}
queryID = campaign.QueryID
// We do not want to use the outer `ctx` directly because we want to cleanup the campaign
// even if the outer `ctx` is canceled (e.g. a client terminating the connection).
// Also, we make sure stats and activity DB operations don't get killed after we return results.
ctxWithoutCancel := context.WithoutCancel(ctx)
defer func() {
err := svc.CompleteCampaign(ctxWithoutCancel, campaign)
if err != nil {
level.Error(svc.logger).Log(
"msg", "completing campaign (sync)", "query.id", campaign.QueryID, "campaign.id", campaign.ID, "err", err,
)
resultsCh <- fleet.QueryCampaignResult{
QueryID: queryID,
Error: ptr.String(err.Error()),
Err: err,
}
}
}()
readChan, cancelFunc, err := svc.GetCampaignReader(ctx, campaign)
if err != nil {
level.Error(svc.logger).Log(
"msg", "get campaign reader", "query.id", campaign.QueryID, "campaign.id", campaign.ID, "err", err,
)
resultsCh <- fleet.QueryCampaignResult{QueryID: queryID, Error: ptr.String(err.Error()), Err: err}
return
}
defer cancelFunc()
var results []fleet.QueryResult
timeout := time.After(deadline)
// We process stats along with results as they are sent back to the user.
// We do a batch update of the stats.
// We update aggregated stats once online hosts have reported.
const statsBatchSize = 1000
perfStatsTracker := statsTracker{}
perfStatsTracker.saveStats, err = svc.ds.IsSavedQuery(ctx, campaign.QueryID)
if err != nil {
level.Error(svc.logger).Log("msg", "error checking saved query", "query.id", campaign.QueryID, "err", err)
perfStatsTracker.saveStats = false
}
totalHosts := campaign.Metrics.TotalHosts
// We update aggregated stats and activity at the end asynchronously.
defer func() {
go func() {
svc.updateStats(ctxWithoutCancel, queryID, svc.logger, &perfStatsTracker, true)
svc.addLiveQueryActivity(ctxWithoutCancel, totalHosts, queryID, svc.logger)
}()
}()
loop:
for {
select {
case res := <-readChan:
switch res := res.(type) {
case fleet.DistributedQueryResult:
results = append(results, fleet.QueryResult{HostID: res.Host.ID, Rows: res.Rows, Error: res.Error})
counterMutex.Lock()
respondedHostIDs[res.Host.ID] = struct{}{}
counterMutex.Unlock()
if perfStatsTracker.saveStats && res.Stats != nil {
perfStatsTracker.stats = append(
perfStatsTracker.stats,
statsToSave{
hostID: res.Host.ID, Stats: res.Stats, outputSize: calculateOutputSize(&perfStatsTracker, &res),
},
)
if len(perfStatsTracker.stats) >= statsBatchSize {
svc.updateStats(ctx, campaign.QueryID, svc.logger, &perfStatsTracker, false)
}
}
if len(results) == len(hostIDs) {
break loop
}
case error:
resultsCh <- fleet.QueryCampaignResult{QueryID: queryID, Error: ptr.String(res.Error()), Err: res}
return
}
case <-timeout:
break loop
case <-ctx.Done():
break loop
}
}
resultsCh <- fleet.QueryCampaignResult{QueryID: queryID, Results: results}
}()
}
// Iterate collecting results until all the goroutines have returned
go func() {
wg.Wait()
close(resultsCh)
}()
var results []fleet.QueryCampaignResult
for result := range resultsCh {
results = append(results, result)
}
return results, len(respondedHostIDs), nil
}
func (svc *Service) GetCampaignReader(ctx context.Context, campaign *fleet.DistributedQueryCampaign) (<-chan interface{}, context.CancelFunc, error) {
// Open the channel from which we will receive incoming query results
// (probably from the redis pubsub implementation)
cancelCtx, cancelFunc := context.WithCancel(ctx)
readChan, err := svc.resultStore.ReadChannel(cancelCtx, *campaign)
if err != nil {
cancelFunc()
return nil, nil, fmt.Errorf("cannot open read channel for campaign %d ", campaign.ID)
}
campaign.Status = fleet.QueryRunning
if err := svc.ds.SaveDistributedQueryCampaign(ctx, campaign); err != nil {
cancelFunc()
return nil, nil, ctxerr.Wrap(ctx, err, "error saving campaign state")
}
return readChan, cancelFunc, nil
}
func (svc *Service) CompleteCampaign(ctx context.Context, campaign *fleet.DistributedQueryCampaign) error {
campaign.Status = fleet.QueryComplete
err := svc.ds.SaveDistributedQueryCampaign(ctx, campaign)
if err != nil {
return ctxerr.Wrap(ctx, err, "saving distributed campaign after complete")
}
err = svc.liveQueryStore.StopQuery(strconv.Itoa(int(campaign.ID)))
if err != nil {
return ctxerr.Wrap(ctx, err, "stopping query after after complete")
}
return nil
}