mirror of
https://gitee.com/milvus-io/milvus.git
synced 2025-12-06 09:08:43 +08:00
fix: Negative -1 executing compaction tasks (#39954)
See also: #39675 --------- Signed-off-by: yangxuan <xuan.yang@zilliz.com>
This commit is contained in:
parent
84df80b5e4
commit
315cfb7f32
@ -722,7 +722,7 @@ func (c *compactionPlanHandler) checkCompaction(assigner NodeAssigner) error {
|
||||
if ok := assigner.assign(t); !ok {
|
||||
break
|
||||
}
|
||||
metrics.DataCoordCompactionTaskNum.WithLabelValues(fmt.Sprintf("%d", NullNodeID), t.GetTaskProto().GetType().String(), metrics.Executing).Dec()
|
||||
metrics.DataCoordCompactionTaskNum.WithLabelValues(fmt.Sprintf("%d", NullNodeID), t.GetTaskProto().GetType().String(), metrics.Pending).Dec()
|
||||
metrics.DataCoordCompactionTaskNum.WithLabelValues(fmt.Sprintf("%d", t.GetTaskProto().GetNodeID()), t.GetTaskProto().GetType().String(), metrics.Executing).Inc()
|
||||
}
|
||||
}
|
||||
|
||||
@ -700,12 +700,18 @@ func (t *clusteringCompactionTask) doCompact() error {
|
||||
}
|
||||
err = t.sessions.Compaction(context.Background(), t.GetTaskProto().GetNodeID(), t.GetPlan())
|
||||
if err != nil {
|
||||
if errors.Is(err, merr.ErrDataNodeSlotExhausted) {
|
||||
log.Warn("fail to notify compaction tasks to DataNode because the node slots exhausted")
|
||||
return t.updateAndSaveTaskMeta(setNodeID(NullNodeID))
|
||||
originNodeID := t.GetTaskProto().GetNodeID()
|
||||
log.Warn("Failed to notify compaction tasks to DataNode",
|
||||
zap.Int64("planID", t.GetTaskProto().GetPlanID()),
|
||||
zap.Int64("nodeID", originNodeID),
|
||||
zap.Error(err))
|
||||
err := t.updateAndSaveTaskMeta(setState(datapb.CompactionTaskState_pipelining), setNodeID(NullNodeID))
|
||||
if err != nil {
|
||||
log.Warn("updateAndSaveTaskMeta fail", zap.Int64("planID", t.GetTaskProto().GetPlanID()), zap.Error(err))
|
||||
return err
|
||||
}
|
||||
log.Warn("Failed to notify compaction tasks to DataNode", zap.Error(err))
|
||||
return t.updateAndSaveTaskMeta(setState(datapb.CompactionTaskState_pipelining), setNodeID(NullNodeID))
|
||||
metrics.DataCoordCompactionTaskNum.WithLabelValues(fmt.Sprintf("%d", originNodeID), t.GetTaskProto().GetType().String(), metrics.Executing).Dec()
|
||||
metrics.DataCoordCompactionTaskNum.WithLabelValues(fmt.Sprintf("%d", NullNodeID), t.GetTaskProto().GetType().String(), metrics.Pending).Inc()
|
||||
}
|
||||
return t.updateAndSaveTaskMeta(setState(datapb.CompactionTaskState_executing))
|
||||
}
|
||||
|
||||
@ -32,6 +32,7 @@ import (
|
||||
"github.com/milvus-io/milvus/internal/datacoord/session"
|
||||
"github.com/milvus-io/milvus/pkg/v2/common"
|
||||
"github.com/milvus-io/milvus/pkg/v2/log"
|
||||
"github.com/milvus-io/milvus/pkg/v2/metrics"
|
||||
"github.com/milvus-io/milvus/pkg/v2/proto/datapb"
|
||||
"github.com/milvus-io/milvus/pkg/v2/util/merr"
|
||||
"github.com/milvus-io/milvus/pkg/v2/util/paramtable"
|
||||
@ -103,8 +104,18 @@ func (t *l0CompactionTask) processPipelining() bool {
|
||||
|
||||
err = t.sessions.Compaction(context.TODO(), t.GetTaskProto().GetNodeID(), plan)
|
||||
if err != nil {
|
||||
log.Warn("l0CompactionTask failed to notify compaction tasks to DataNode", zap.Int64("planID", t.GetTaskProto().GetPlanID()), zap.Error(err))
|
||||
t.updateAndSaveTaskMeta(setState(datapb.CompactionTaskState_pipelining), setNodeID(NullNodeID))
|
||||
originNodeID := t.GetTaskProto().GetNodeID()
|
||||
log.Warn("l0CompactionTask failed to notify compaction tasks to DataNode",
|
||||
zap.Int64("planID", t.GetTaskProto().GetPlanID()),
|
||||
zap.Int64("nodeID", originNodeID),
|
||||
zap.Error(err))
|
||||
err := t.updateAndSaveTaskMeta(setState(datapb.CompactionTaskState_pipelining), setNodeID(NullNodeID))
|
||||
if err != nil {
|
||||
log.Warn("l0CompactionTask failed to updateAndSaveTaskMeta", zap.Int64("planID", t.GetTaskProto().GetPlanID()), zap.Error(err))
|
||||
return false
|
||||
}
|
||||
metrics.DataCoordCompactionTaskNum.WithLabelValues(fmt.Sprintf("%d", originNodeID), t.GetTaskProto().GetType().String(), metrics.Executing).Dec()
|
||||
metrics.DataCoordCompactionTaskNum.WithLabelValues(fmt.Sprintf("%d", NullNodeID), t.GetTaskProto().GetType().String(), metrics.Pending).Inc()
|
||||
return false
|
||||
}
|
||||
|
||||
|
||||
@ -14,6 +14,7 @@ import (
|
||||
"github.com/milvus-io/milvus/internal/datacoord/allocator"
|
||||
"github.com/milvus-io/milvus/internal/datacoord/session"
|
||||
"github.com/milvus-io/milvus/pkg/v2/log"
|
||||
"github.com/milvus-io/milvus/pkg/v2/metrics"
|
||||
"github.com/milvus-io/milvus/pkg/v2/proto/datapb"
|
||||
"github.com/milvus-io/milvus/pkg/v2/util/merr"
|
||||
"github.com/milvus-io/milvus/pkg/v2/util/paramtable"
|
||||
@ -70,11 +71,17 @@ func (t *mixCompactionTask) processPipelining() bool {
|
||||
// Compaction tasks may be refused by DataNode because of slot limit. In this case, the node id is reset
|
||||
// to enable a retry in compaction.checkCompaction().
|
||||
// This is tricky, we should remove the reassignment here.
|
||||
log.Warn("mixCompactionTask failed to notify compaction tasks to DataNode", zap.Int64("planID", t.GetTaskProto().GetPlanID()), zap.Error(err))
|
||||
originNodeID := t.GetTaskProto().GetNodeID()
|
||||
log.Warn("mixCompactionTask failed to notify compaction tasks to DataNode",
|
||||
zap.Int64("planID", t.GetTaskProto().GetPlanID()),
|
||||
zap.Int64("nodeID", originNodeID),
|
||||
zap.Error(err))
|
||||
err = t.updateAndSaveTaskMeta(setState(datapb.CompactionTaskState_pipelining), setNodeID(NullNodeID))
|
||||
if err != nil {
|
||||
log.Warn("mixCompactionTask failed to updateAndSaveTaskMeta", zap.Error(err))
|
||||
}
|
||||
metrics.DataCoordCompactionTaskNum.WithLabelValues(fmt.Sprintf("%d", originNodeID), t.GetTaskProto().GetType().String(), metrics.Executing).Dec()
|
||||
metrics.DataCoordCompactionTaskNum.WithLabelValues(fmt.Sprintf("%d", NullNodeID), t.GetTaskProto().GetType().String(), metrics.Pending).Inc()
|
||||
return false
|
||||
}
|
||||
log.Info("mixCompactionTask notify compaction tasks to DataNode")
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user