ClusterCockpit/cc-backend
1
0
Fork 0
mirror of https://github.com/ClusterCockpit/cc-backend synced 2024-12-26 13:29:05 +01:00
Code Issues Packages Releases Activity

improve transaction init error handling

Browse Source
This commit is contained in:
Christoph Kluge 2024-11-25 17:03:59 +01:00
parent d4f487d554
commit 8ea1454c06
1 changed files with 19 additions and 22 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

25
internal/taskManager/updateFootprintService.go
View File

@ -43,6 +43,9 @@ func RegisterFootprintWorker() {
if err != nil { if err != nil {
continue continue
} }
// NOTE: Additional Subcluster Loop Could Allow For Limited List Of (Energy)Footprint-Metrics Only.
// - Chunk-Size Would Then Be 'SubCluster' (Running Jobs, Transactions) as Lists Can Change Within SCs
// - Would Require Review of 'updateFootprint' And 'updateEnergy' Usage
allMetrics := make([]string, 0) allMetrics := make([]string, 0)
metricConfigs := archive.GetCluster(cluster.Name).MetricConfig metricConfigs := archive.GetCluster(cluster.Name).MetricConfig
for _, mc := range metricConfigs { for _, mc := range metricConfigs {
@ -78,7 +81,7 @@ func RegisterFootprintWorker() {
for _, metric := range allMetrics { for _, metric := range allMetrics {
avg, min, max := 0.0, 0.0, 0.0 avg, min, max := 0.0, 0.0, 0.0
data, ok := jobStats[metric] // Metric:[Hostname:Stats] data, ok := jobStats[metric] // JobStats[Metric1:[Hostname1:[Stats], Hostname2:[Stats], ...], Metric2[...] ...]
if ok { if ok {
for _, res := range job.Resources { for _, res := range job.Resources {
hostStats, ok := data[res.Hostname] hostStats, ok := data[res.Hostname]
@ -87,14 +90,9 @@ func RegisterFootprintWorker() {
min = math.Min(min, hostStats.Min) min = math.Min(min, hostStats.Min)
max = math.Max(max, hostStats.Max) max = math.Max(max, hostStats.Max)
} }
// else {
// log.Debugf("no stats data return for host %s in job %d, metric %s", res.Hostname, job.JobID, metric)
// }
} }
} }
// else {
// log.Debugf("no stats data return for job %d, metric %s", job.JobID, metric)
// }
// Add values rounded to 2 digits // Add values rounded to 2 digits
jobMeta.Statistics[metric] = schema.JobStatistics{ jobMeta.Statistics[metric] = schema.JobStatistics{
@ -131,23 +129,22 @@ func RegisterFootprintWorker() {
t, err := jobRepo.TransactionInit() t, err := jobRepo.TransactionInit()
if err != nil { if err != nil {
log.Errorf("failed TransactionInit %v", err) log.Errorf("failed TransactionInit %v", err)
} log.Errorf("skipped %d transactions for cluster %s", len(pendingStatements), cluster.Name)
ce += len(pendingStatements)
for idx, ps := range pendingStatements { } else {
for _, ps := range pendingStatements {
query, args, err := ps.ToSql() query, args, err := ps.ToSql()
if err != nil { if err != nil {
log.Errorf("failed in ToSQL conversion: %v", err) log.Errorf("failed in ToSQL conversion: %v", err)
ce++ ce++
} else { } else {
// Args: JSON, JSON, ENERGY, JOBID // args...: Footprint-JSON, Energyfootprint-JSON, TotalEnergy, JobID
log.Debugf("add transaction on index %d", idx)
jobRepo.TransactionAdd(t, query, args...) jobRepo.TransactionAdd(t, query, args...)
c++ c++
} }
} }
jobRepo.TransactionEnd(t) jobRepo.TransactionEnd(t)
}
log.Debugf("Finish Cluster %s, took %s", cluster.Name, time.Since(s_cluster)) log.Debugf("Finish Cluster %s, took %s", cluster.Name, time.Since(s_cluster))
} }
log.Printf("Updating %d (of %d; Skipped %d) Footprints is done and took %s", c, cl, ce, time.Since(s)) log.Printf("Updating %d (of %d; Skipped %d) Footprints is done and took %s", c, cl, ce, time.Since(s))