github/containerd
2
0
Fork 0
Code Issues Pull Requests Packages Projects Releases 2 Wiki Activity

add metrics for discarding events

Browse Source 
Signed-off-by: rongfu.leng <rongfu.leng@daocloud.io>
This commit is contained in:
rongfu.leng 2023-08-08 22:40:27 +08:00
parent 165f8e414e
commit 54baf766e5
4 changed files with 16 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
pkg/cri/sbserver/helpers.go
View File

@ -457,6 +457,7 @@ func (c *criService) generateAndSendContainerEvent(ctx context.Context, containe
select { select {
case c.containerEventsChan <- event: case c.containerEventsChan <- event:
default: default:
containerEventsDroppedCount.Inc()
log.G(ctx).Debugf("containerEventsChan is full, discarding event %+v", event) log.G(ctx).Debugf("containerEventsChan is full, discarding event %+v", event)
} }
} }

12
pkg/cri/sbserver/metrics.go
View File

@ -29,11 +29,12 @@ var (
sandboxRuntimeStopTimer metrics.LabeledTimer sandboxRuntimeStopTimer metrics.LabeledTimer
sandboxRemoveTimer metrics.LabeledTimer sandboxRemoveTimer metrics.LabeledTimer
containerListTimer metrics.Timer containerListTimer metrics.Timer
containerRemoveTimer metrics.LabeledTimer containerRemoveTimer metrics.LabeledTimer
containerCreateTimer metrics.LabeledTimer containerCreateTimer metrics.LabeledTimer
containerStopTimer metrics.LabeledTimer containerStopTimer metrics.LabeledTimer
containerStartTimer metrics.LabeledTimer containerStartTimer metrics.LabeledTimer
containerEventsDroppedCount metrics.Counter
networkPluginOperations metrics.LabeledCounter networkPluginOperations metrics.LabeledCounter
networkPluginOperationsErrors metrics.LabeledCounter networkPluginOperationsErrors metrics.LabeledCounter
@ -57,6 +58,7 @@ func init() {
containerCreateTimer = ns.NewLabeledTimer("container_create", "time to create a container", "runtime") containerCreateTimer = ns.NewLabeledTimer("container_create", "time to create a container", "runtime")
containerStopTimer = ns.NewLabeledTimer("container_stop", "time to stop a container", "runtime") containerStopTimer = ns.NewLabeledTimer("container_stop", "time to stop a container", "runtime")
containerStartTimer = ns.NewLabeledTimer("container_start", "time to start a container", "runtime") containerStartTimer = ns.NewLabeledTimer("container_start", "time to start a container", "runtime")
containerEventsDroppedCount = ns.NewCounter("container_events_dropped", "count container discarding event total from server start")
networkPluginOperations = ns.NewLabeledCounter("network_plugin_operations_total", "cumulative number of network plugin operations by operation type", "operation_type") networkPluginOperations = ns.NewLabeledCounter("network_plugin_operations_total", "cumulative number of network plugin operations by operation type", "operation_type")
networkPluginOperationsErrors = ns.NewLabeledCounter("network_plugin_operations_errors_total", "cumulative number of network plugin operations by operation type", "operation_type") networkPluginOperationsErrors = ns.NewLabeledCounter("network_plugin_operations_errors_total", "cumulative number of network plugin operations by operation type", "operation_type")

1
pkg/cri/server/helpers.go
View File

@ -534,6 +534,7 @@ func (c *criService) generateAndSendContainerEvent(ctx context.Context, containe
select { select {
case c.containerEventsChan <- event: case c.containerEventsChan <- event:
default: default:
containerEventsDroppedCount.Inc()
logrus.Debugf("containerEventsChan is full, discarding event %+v", event) logrus.Debugf("containerEventsChan is full, discarding event %+v", event)
} }
} }

12
pkg/cri/server/metrics.go
View File

@ -30,11 +30,12 @@ var (
sandboxRuntimeStopTimer metrics.LabeledTimer sandboxRuntimeStopTimer metrics.LabeledTimer
sandboxRemoveTimer metrics.LabeledTimer sandboxRemoveTimer metrics.LabeledTimer
containerListTimer metrics.Timer containerListTimer metrics.Timer
containerRemoveTimer metrics.LabeledTimer containerRemoveTimer metrics.LabeledTimer
containerCreateTimer metrics.LabeledTimer containerCreateTimer metrics.LabeledTimer
containerStopTimer metrics.LabeledTimer containerStopTimer metrics.LabeledTimer
containerStartTimer metrics.LabeledTimer containerStartTimer metrics.LabeledTimer
containerEventsDroppedCount metrics.Counter
networkPluginOperations metrics.LabeledCounter networkPluginOperations metrics.LabeledCounter
networkPluginOperationsErrors metrics.LabeledCounter networkPluginOperationsErrors metrics.LabeledCounter
@ -68,6 +69,7 @@ func init() {
containerCreateTimer = ns.NewLabeledTimer("container_create", "time to create a container", "runtime") containerCreateTimer = ns.NewLabeledTimer("container_create", "time to create a container", "runtime")
containerStopTimer = ns.NewLabeledTimer("container_stop", "time to stop a container", "runtime") containerStopTimer = ns.NewLabeledTimer("container_stop", "time to stop a container", "runtime")
containerStartTimer = ns.NewLabeledTimer("container_start", "time to start a container", "runtime") containerStartTimer = ns.NewLabeledTimer("container_start", "time to start a container", "runtime")
containerEventsDroppedCount = ns.NewCounter("container_events_dropped", "count container discarding event total from server start")
networkPluginOperations = ns.NewLabeledCounter("network_plugin_operations_total", "cumulative number of network plugin operations by operation type", "operation_type") networkPluginOperations = ns.NewLabeledCounter("network_plugin_operations_total", "cumulative number of network plugin operations by operation type", "operation_type")
networkPluginOperationsErrors = ns.NewLabeledCounter("network_plugin_operations_errors_total", "cumulative number of network plugin operations by operation type", "operation_type") networkPluginOperationsErrors = ns.NewLabeledCounter("network_plugin_operations_errors_total", "cumulative number of network plugin operations by operation type", "operation_type")