JakeHillion/storj
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

satellite/repair: add monitoring for remote_segments_healthy_percentage

Browse Source 
Change-Id: I6ad29fe1a947ac19d15e40ea33164a510eb33d4f
This commit is contained in:
paul cannon 2020-03-10 08:59:29 -05:00 committed by Maximillian von Briesen
parent 2f991b6c56
commit ba5991dc86
2 changed files with 10 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
monkit.lock
View File

@ -60,6 +60,8 @@ storj.io/storj/satellite/repair/checker."checker_segment_total_count" IntVal
storj.io/storj/satellite/repair/checker."remote_files_checked" IntVal
storj.io/storj/satellite/repair/checker."remote_files_lost" IntVal
storj.io/storj/satellite/repair/checker."remote_segments_checked" IntVal
storj.io/storj/satellite/repair/checker."remote_segments_failed_to_check" IntVal
storj.io/storj/satellite/repair/checker."remote_segments_healthy_percentage" FloatVal
storj.io/storj/satellite/repair/checker."remote_segments_lost" IntVal
storj.io/storj/satellite/repair/checker."remote_segments_needing_repair" IntVal
storj.io/storj/satellite/repair/checker."remote_segments_over_threshold_1" IntVal

8
satellite/repair/checker/checker.go
View File

@ -43,6 +43,7 @@ type durabilityStats struct {
remoteSegmentsChecked int64
remoteSegmentsNeedingRepair int64
remoteSegmentsLost int64
remoteSegmentsFailedToCheck int64
remoteSegmentInfo []string
// remoteSegmentsOverThreshold[0]=# of healthy=rt+1, remoteSegmentsOverThreshold[1]=# of healthy=rt+2, etc...
remoteSegmentsOverThreshold [5]int64
@ -130,6 +131,7 @@ func (checker *Checker) IdentifyInjuredSegments(ctx context.Context) (err error)
mon.IntVal("remote_files_checked").Observe(observer.monStats.objectsChecked) //locked
mon.IntVal("remote_segments_checked").Observe(observer.monStats.remoteSegmentsChecked) //locked
mon.IntVal("remote_segments_failed_to_check").Observe(observer.monStats.remoteSegmentsFailedToCheck) //locked
mon.IntVal("remote_segments_needing_repair").Observe(observer.monStats.remoteSegmentsNeedingRepair) //locked
mon.IntVal("remote_segments_lost").Observe(observer.monStats.remoteSegmentsLost) //locked
mon.IntVal("remote_files_lost").Observe(int64(len(observer.monStats.remoteSegmentInfo))) //locked
@ -139,6 +141,11 @@ func (checker *Checker) IdentifyInjuredSegments(ctx context.Context) (err error)
mon.IntVal("remote_segments_over_threshold_4").Observe(observer.monStats.remoteSegmentsOverThreshold[3]) //locked
mon.IntVal("remote_segments_over_threshold_5").Observe(observer.monStats.remoteSegmentsOverThreshold[4]) //locked
allUnhealthy := observer.monStats.remoteSegmentsNeedingRepair + observer.monStats.remoteSegmentsFailedToCheck
allChecked := observer.monStats.remoteSegmentsChecked
allHealthy := allChecked - allUnhealthy
mon.FloatVal("remote_segments_healthy_percentage").Observe(100 * float64(allHealthy) / float64(allChecked)) //locked
return nil
}
@ -245,6 +252,7 @@ func (obs *checkerObserver) RemoteSegment(ctx context.Context, path metainfo.Sco
missingPieces, err := obs.nodestate.MissingPieces(ctx, pointer.CreationDate, pieces)
if err != nil {
obs.monStats.remoteSegmentsFailedToCheck++
return errs.Combine(Error.New("error getting missing pieces"), err)
}