TiCDC Alert Rules

This document describes the TiCDC alert rules and the corresponding solutions. In descending order, the severity levels are: Critical, Warning.

Critical alerts

This section introduces critical alerts and solutions.

cdc_checkpoint_high_delay

For critical alerts, you need to pay close attention to abnormal monitoring metrics.

• Alert rule:

  (time() - ticdc_owner_checkpoint_ts / 1000) > 600

• Description:

  A replication task is delayed more than 10 minutes.

• Solution:

  See TiCDC Handles Replication Interruption.

cdc_resolvedts_high_delay

• Alert rule:

  (time() - ticdc_owner_resolved_ts / 1000) > 300

• Description:

  The Resolved TS of a replication task is delayed more than 5 minutes.

• Solution:

  See TiCDC Handles Replication Interruption.

ticdc_changefeed_failed

• Alert rule:

  (max_over_time(ticdc_owner_status[1m]) == 2) > 0

• Description:

  A replication task encounters an unrecoverable error and enters the failed state.

• Solution:

  This alert is similar to replication interruption. See TiCDC Handles Replication Interruption.

ticdc_processor_exit_with_error_count

• Alert rule:

  changes(ticdc_processor_exit_with_error_count[1m]) > 0

• Description:

  A replication task reports an error and exits.

• Solution:

  See TiCDC Handles Replication Interruption.

Warning alerts

Warning alerts are a reminder for an issue or error.

cdc_multiple_owners

• Alert rule:

  sum(rate(ticdc_owner_ownership_counter[30s])) >= 2

• Description:

  There are multiple owners in the TiCDC cluster.

• Solution:

  Collect TiCDC logs to locate the root cause.

cdc_sink_flush_duration_time_more_than_10s

• Alert rule:

  histogram_quantile(0.9, rate(ticdc_sink_txn_worker_flush_duration[1m])) > 10

• Description:

  It takes a replication task more than 10 seconds to write data to the downstream database.

• Solution:

  Check whether there are problems in the downstream database.

cdc_processor_checkpoint_tso_no_change_for_1m

• Alert rule:

  changes(ticdc_processor_checkpoint_ts[1m]) < 1

• Description:

  A replication task has not advanced for more than 1 minute.

• Solution:

  See TiCDC Handles Replication Interruption.

ticdc_puller_entry_sorter_sort_bucket

• Alert rule:

  histogram_quantile(0.9, rate(ticdc_puller_entry_sorter_sort_bucket{}[1m])) > 1

• Description:

  The delay of TiCDC puller entry sorter is too high.

• Solution:

  Collect TiCDC logs to locate the root cause.

ticdc_puller_entry_sorter_merge_bucket

• Alert rule:

  histogram_quantile(0.9, rate(ticdc_puller_entry_sorter_merge_bucket{}[1m])) > 1

• Description:

  The delay of TiCDC puller entry sorter merge is too high.

• Solution:

  Collect TiCDC logs to locate the root cause.

tikv_cdc_min_resolved_ts_no_change_for_1m

• Alert rule:

  changes(tikv_cdc_min_resolved_ts[1m]) < 1 and ON (instance) tikv_cdc_region_resolve_status{status="resolved"} > 0 and ON (instance) tikv_cdc_captured_region_total > 0

• Description:

  The minimum Resolved TS 1 of TiKV CDC has not advanced for 1 minute.

• Solution:

  Collect TiKV logs to locate the root cause.

tikv_cdc_scan_duration_seconds_more_than_10min

• Alert rule:

  histogram_quantile(0.9, rate(tikv_cdc_scan_duration_seconds_bucket{}[1m])) > 600

• Description:

  The TiKV CDC module has scanned for incremental replication for more than 10 minutes.

• Solution:

  Collect TiCDC monitoring metrics and TiKV logs to locate the root cause.

ticdc_sink_mysql_execution_error

• Alert rule:

  changes(ticdc_sink_mysql_execution_error[1m]) > 0

• Description:

  An error occurs when a replication task writes data to the downstream MySQL.

• Solution:

  There are many possible root causes. See Troubleshoot TiCDC.

ticdc_memory_abnormal

• Alert rule:

  go_memstats_heap_alloc_bytes{job="ticdc"} > 1e+10

• Description:

  The TiCDC heap memory usage exceeds 10 GiB.

• Solution:

  Collect TiCDC logs to locate the root cause.