DWS_2000000023 Vacuum Full Operation That Holds a Table Lock for a Long Time Exists in the DWS Cluster
Alarm Description
VACUUM FULL holds a level-8 lock on a table. If it holds the lock on a table for longer than 20 minutes (or another user-defined value), a major alarm is reported, indicating that the VACUUM FULL operation holds a lock for too long in the cluster. This major alarm is cleared when VACUUM FULL is complete.
Attributes
| Alarm ID | Alarm Category | Alarm Severity | Alarm Type | Service Type | Auto Cleared |
|---|---|---|---|---|---|
| DWS_2000000023 | Tenant plane | Important | Operation alarm | DWS | Yes |
Alarm Parameters
| Category | Name | Description |
|---|---|---|
| Location information | Name | A Vacuum Full Operation That Holds a Table Lock for A Long Time Exists in the Cluster |
| Type | Operation alarm | |
| Generation time | Time when the alarm is generated | |
| Other information | Cluster ID | Cluster details such as resourceId and domain_id |
Alarm Metrics
Call the Querying Collected Data of a Specified Metric - ListMetricsData API to obtain the alarm metric. The metric name in the API is SessionStats.
- Example request:
GET https://{Endpoint}/v1/{project_id}/clusters/{cluster_id}/dms/metrics/SessionStats?offset=0&limit=1&from=1786782138946&to=1786785551018&order_by=ctime&sort_by=desc - Example response:
The alarm metric is lock_mode whose value is not -, lock_relname whose value is not -, and stmt_type whose value is VACUUM in the response. lock_mode indicates the lock mode, lock_relname indicates the resource that is waiting for the lock, and stmt_type indicates the statement type.
{ "code": 0, "msg": "OK", "data": [ { "virtual_cluster_id": 81882, "waiting": false, "query_id": "0", "backend_start": 1786782298, "user_name": "Ruby", "pid": "140504051969432", "resource_pool": "default_pool", "client_port": "20482", "application_name": "wlm", "elapsed_time": 11, "ctime": 1786782308000, "curr_xact_start": -1, "client_address": "172.16.114.248", "state": "idle", "lock_type": "-", "lock_relname": "-", "stmt_type": "COMMIT PREPARED", "query_elapsed_time": -1, "control_group": "", "query": "-", "query_start": 1786782298, "db_name": "gaussdb", "client_hostname": "", "system_query": true, "lock_mode": "-", "inst_name": "cn_5001", "state_change": 1786782298 } ], "count": 743 }
Impact on the System
Other operations fail to hold the lock on the table. As a result, workloads cannot be executed.
Possible Causes
There is a VACUUM FULL operation that holds a table lock for a long time in the cluster.
Handling Procedure
- In the navigation pane of the monitoring panel, choose Monitoring > Real-Time Queries > Sessions. In the session list, set the search criteria to LIKE and search for the keyword vacuum full.

- Check whether there is a table lock waiting for VACUUM FULL to complete by querying the locked object.

- Check whether the VACUUM FULL operation needs to be handled.
- Check whether VACUUM FULL is a system behavior and whether it affects system functions. If VACUUM FULL does not affect other service queries, wait until it is complete. The alarm will be automatically cleared.
- If VACUUM FULL affects normal service execution, you can find and kill related sessions on the Real-Time Queries tab and re-execute VACUUM FULL later.
Alarm Clearance
This alarm is automatically cleared when the VACUUM FULL operation is complete.
What is your overall rating for this page?
Thank you very much for your feedback. We will continue working to improve the documentation.See the reply and handling status in My Cloud VOC.
For any further questions, feel free to contact us through the chatbot.
Chatbot