[Bug] discrepancies between auditlog and profile reported numbers #51646
Open
Description
Steps to reproduce the behavior (Required)
understand discrepancies between auditlog and profile reported numbers:
Select query:
------------
From auditlog:
CpuCostNs
1915775920858
31.93 min
does not match anything in profile
MemCostBytes
12545335920
11.68 GB
matches QueryPeakMemoryUsagePerNode
Time
2336
2.336 sec
matches QueryExecutionWallTime
From profile:
Execution:
- FrontendProfileMergeTime: 79.260ms
- QueryAllocatedMemoryUsage: 2.590 TB
- QueryCumulativeCpuTime: 1h54m
- QueryCumulativeNetworkTime: 196.241ms
- QueryCumulativeOperatorTime: 8s131ms
- QueryCumulativeScanTime: 2s256ms
- QueryDeallocatedMemoryUsage: 2.589 TB
- QueryExecutionWallTime: 2s324ms
- QueryPeakMemoryUsagePerNode: 11.684 GB
- QueryPeakScheduleTime: 944.846ms
- QuerySpillBytes: 0.000 B
- QuerySumMemoryUsage: 873.853 GB
- ResultDeliverTime: 0ns
Update query
------------
auditlog:
cpuCostNs
39228243805444
10.9h
matches QueryCumulativeCpuTime
memCostBytes
1048626061752
976.61GiB
matches QuerySumMemoryUsage
Time
114874
114.87s
does not match with profile
Execution:
- FrontendProfileMergeTime: 21.210ms
- QueryAllocatedMemoryUsage: 9.067 TB
- QueryCumulativeCpuTime: 10h53m
- QueryCumulativeNetworkTime: 0ns
- QueryCumulativeOperatorTime: 1m40s
- QueryCumulativeScanTime: 0ns
- QueryDeallocatedMemoryUsage: 9.052 TB
- QueryExecutionWallTime: 1m48s
- QueryPeakMemoryUsagePerNode: 52.847 GB
- QueryPeakScheduleTime: 57.050ms
- QuerySpillBytes: 0.000 B
- QuerySumMemoryUsage: 976.609 GB
- ResultDeliverTime: 2s827ms
Expected behavior (Required)
Introduce two metrics: peak
Real behavior (Required)
Make the metrics from audit log consistent with the metrics of the profile
StarRocks version (Required)
N/A