mirror of
https://github.com/pezkuwichain/pezkuwi-subxt.git
synced 2026-08-02 02:55:40 +00:00
Max Inden
fe76ebd548
* .maintain/monitoring: Add alerting rule tests * .maintain/monitoring/alerting-rules/alerting-rules.yaml: Break lines * .gitlab-ci.yml: Add promtool rule testing step
240 lines
9.4 KiB
YAML
240 lines
9.4 KiB
YAML
rule_files:
|
|
- /dev/stdin
|
|
|
|
evaluation_interval: 1m
|
|
|
|
tests:
|
|
- interval: 1m
|
|
input_series:
|
|
- series: 'polkadot_sub_libp2p_peers_count{
|
|
job="polkadot",
|
|
pod="polkadot-abcdef01234-abcdef",
|
|
instance="polkadot-abcdef01234-abcdef",
|
|
}'
|
|
values: '3 2+0x4 1+0x9' # 3 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1
|
|
|
|
- series: 'polkadot_sub_txpool_validations_scheduled{
|
|
job="polkadot",
|
|
pod="polkadot-abcdef01234-abcdef",
|
|
instance="polkadot-abcdef01234-abcdef",
|
|
}'
|
|
values: '10+1x30' # 10 11 12 13 .. 40
|
|
|
|
- series: 'polkadot_sub_txpool_validations_finished{
|
|
job="polkadot",
|
|
pod="polkadot-abcdef01234-abcdef",
|
|
instance="polkadot-abcdef01234-abcdef",
|
|
}'
|
|
values: '0x30' # 0 0 0 0 .. 0
|
|
|
|
- series: 'polkadot_block_height{
|
|
status="best", job="polkadot",
|
|
pod="polkadot-abcdef01234-abcdef",
|
|
instance="polkadot-abcdef01234-abcdef",
|
|
}'
|
|
values: '1+1x3 4+0x13' # 1 2 3 4 4 4 4 4 4 4 4 4 ...
|
|
|
|
- series: 'polkadot_block_height{
|
|
status="finalized",
|
|
job="polkadot",
|
|
pod="polkadot-abcdef01234-abcdef",
|
|
instance="polkadot-abcdef01234-abcdef",
|
|
}'
|
|
values: '1+1x3 4+0x13' # 1 2 3 4 4 4 4 4 4 4 4 4 ...
|
|
|
|
- series: 'polkadot_cpu_usage_percentage{
|
|
job="polkadot",
|
|
pod="polkadot-abcdef01234-abcdef",
|
|
instance="polkadot-abcdef01234-abcdef",
|
|
}'
|
|
values: '0+20x5 100+0x5' # 0 20 40 60 80 100 100 100 100 100 100
|
|
|
|
alert_rule_test:
|
|
|
|
######################################################################
|
|
# Resource usage
|
|
######################################################################
|
|
|
|
- eval_time: 9m
|
|
alertname: HighCPUUsage
|
|
exp_alerts:
|
|
- eval_time: 10m
|
|
alertname: HighCPUUsage
|
|
exp_alerts:
|
|
- exp_labels:
|
|
severity: warning
|
|
pod: polkadot-abcdef01234-abcdef
|
|
instance: polkadot-abcdef01234-abcdef
|
|
job: polkadot
|
|
exp_annotations:
|
|
message: "The node polkadot-abcdef01234-abcdef has a CPU
|
|
usage higher than 100% for more than 5 minutes"
|
|
|
|
######################################################################
|
|
# Block production
|
|
######################################################################
|
|
|
|
- eval_time: 6m
|
|
alertname: LowNumberOfNewBlocks
|
|
exp_alerts:
|
|
- eval_time: 7m
|
|
alertname: LowNumberOfNewBlocks
|
|
exp_alerts:
|
|
- exp_labels:
|
|
severity: warning
|
|
pod: polkadot-abcdef01234-abcdef
|
|
instance: polkadot-abcdef01234-abcdef
|
|
job: polkadot
|
|
status: best
|
|
exp_annotations:
|
|
message: "Less than one new block per minute on instance
|
|
polkadot-abcdef01234-abcdef."
|
|
|
|
- eval_time: 14m
|
|
alertname: LowNumberOfNewBlocks
|
|
exp_alerts:
|
|
- exp_labels:
|
|
severity: warning
|
|
pod: polkadot-abcdef01234-abcdef
|
|
instance: polkadot-abcdef01234-abcdef
|
|
job: polkadot
|
|
status: best
|
|
exp_annotations:
|
|
message: "Less than one new block per minute on instance
|
|
polkadot-abcdef01234-abcdef."
|
|
- exp_labels:
|
|
severity: critical
|
|
pod: polkadot-abcdef01234-abcdef
|
|
instance: polkadot-abcdef01234-abcdef
|
|
job: polkadot
|
|
status: best
|
|
exp_annotations:
|
|
message: "Less than one new block per minute on instance
|
|
polkadot-abcdef01234-abcdef."
|
|
|
|
######################################################################
|
|
# Block finalization
|
|
######################################################################
|
|
|
|
- eval_time: 6m
|
|
alertname: BlockFinalizationSlow
|
|
exp_alerts:
|
|
- eval_time: 7m
|
|
alertname: BlockFinalizationSlow
|
|
exp_alerts:
|
|
- exp_labels:
|
|
severity: warning
|
|
pod: polkadot-abcdef01234-abcdef
|
|
instance: polkadot-abcdef01234-abcdef
|
|
job: polkadot
|
|
status: finalized
|
|
exp_annotations:
|
|
message: "Finalized block on instance
|
|
polkadot-abcdef01234-abcdef increases by less than 1 per
|
|
minute."
|
|
|
|
- eval_time: 14m
|
|
alertname: BlockFinalizationSlow
|
|
exp_alerts:
|
|
- exp_labels:
|
|
severity: warning
|
|
pod: polkadot-abcdef01234-abcdef
|
|
instance: polkadot-abcdef01234-abcdef
|
|
job: polkadot
|
|
status: finalized
|
|
exp_annotations:
|
|
message: "Finalized block on instance
|
|
polkadot-abcdef01234-abcdef increases by less than 1 per
|
|
minute."
|
|
- exp_labels:
|
|
severity: critical
|
|
pod: polkadot-abcdef01234-abcdef
|
|
instance: polkadot-abcdef01234-abcdef
|
|
job: polkadot
|
|
status: finalized
|
|
exp_annotations:
|
|
message: "Finalized block on instance
|
|
polkadot-abcdef01234-abcdef increases by less than 1 per
|
|
minute."
|
|
|
|
######################################################################
|
|
# Transaction queue
|
|
######################################################################
|
|
|
|
- eval_time: 10m
|
|
alertname: TransactionQueueSize
|
|
exp_alerts:
|
|
- eval_time: 11m
|
|
alertname: TransactionQueueSize
|
|
exp_alerts:
|
|
- exp_labels:
|
|
severity: warning
|
|
pod: polkadot-abcdef01234-abcdef
|
|
instance: polkadot-abcdef01234-abcdef
|
|
job: polkadot
|
|
exp_annotations:
|
|
message: "The node polkadot-abcdef01234-abcdef has more
|
|
than 10 transactions in the queue for more than 10
|
|
minutes"
|
|
|
|
- eval_time: 31m
|
|
alertname: TransactionQueueSize
|
|
exp_alerts:
|
|
- exp_labels:
|
|
severity: warning
|
|
pod: polkadot-abcdef01234-abcdef
|
|
instance: polkadot-abcdef01234-abcdef
|
|
job: polkadot
|
|
exp_annotations:
|
|
message: "The node polkadot-abcdef01234-abcdef has more
|
|
than 10 transactions in the queue for more than 10
|
|
minutes"
|
|
- exp_labels:
|
|
severity: critical
|
|
pod: polkadot-abcdef01234-abcdef
|
|
instance: polkadot-abcdef01234-abcdef
|
|
job: polkadot
|
|
exp_annotations:
|
|
message: "The node polkadot-abcdef01234-abcdef has more
|
|
than 10 transactions in the queue for more than 30
|
|
minutes"
|
|
|
|
######################################################################
|
|
# Networking
|
|
######################################################################
|
|
|
|
- eval_time: 3m # Values: 3 2 2
|
|
alertname: LowNumberOfPeers
|
|
exp_alerts:
|
|
- eval_time: 4m # Values: 2 2 2
|
|
alertname: LowNumberOfPeers
|
|
exp_alerts:
|
|
- exp_labels:
|
|
severity: warning
|
|
pod: polkadot-abcdef01234-abcdef
|
|
instance: polkadot-abcdef01234-abcdef
|
|
job: polkadot
|
|
exp_annotations:
|
|
message: "The node polkadot-abcdef01234-abcdef has less
|
|
than 3 peers for more than 3 minutes"
|
|
|
|
- eval_time: 16m # Values: 3 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1 1
|
|
alertname: LowNumberOfPeers
|
|
exp_alerts:
|
|
- exp_labels:
|
|
severity: warning
|
|
pod: polkadot-abcdef01234-abcdef
|
|
instance: polkadot-abcdef01234-abcdef
|
|
job: polkadot
|
|
exp_annotations:
|
|
message: "The node polkadot-abcdef01234-abcdef has less
|
|
than 3 peers for more than 3 minutes"
|
|
- exp_labels:
|
|
severity: critical
|
|
pod: polkadot-abcdef01234-abcdef
|
|
instance: polkadot-abcdef01234-abcdef
|
|
job: polkadot
|
|
exp_annotations:
|
|
message: "The node polkadot-abcdef01234-abcdef has less
|
|
than 3 peers for more than 15 minutes"
|