Closed zhuwenxing closed 2 years ago
more log info test-e2e-05-05-14-51.zip
in the log file pod_log_previous/milvus-chaos-querynode-96fbf4fdf-g2ncd.log
[2022/05/05 06:34:05.495 +00:00] [DEBUG] [segment.go:313] ["do search on segment"] [segmentID=432991742724145154] [segmentType=3]
[2022/05/05 06:34:05.497 +00:00] [DEBUG] [query_shard.go:238] ["wait serviceable ts done"] [tsType="Delta tSafe"] ["guarantee ts"=432991750143082497] ["serviceable ts"=432991750155665409] [channel=by-dev-rootcoord-dml_3_432991732513636353v1]
[2022/05/05 06:34:05.497 +00:00] [DEBUG] [historical.go:169] ["search validated partitions"] [collectionID=432991732513636353] [partitionIDs="[432991732513636354]"]
[2022/05/05 06:34:05.497 +00:00] [DEBUG] [segment.go:313] ["do search on segment"] [segmentID=432991736511332354] [segmentType=3]
[2022/05/05 06:34:05.497 +00:00] [DEBUG] [segment.go:313] ["do search on segment"] [segmentID=432991736511332353] [segmentType=3]
[2022/05/05 06:34:05.497 +00:00] [DEBUG] [segment.go:313] ["do search on segment"] [segmentID=432991742724145153] [segmentType=3]
[2022/05/05 06:34:05.504 +00:00] [DEBUG] [reduce.go:74] ["start marshal..."] [collectionID=432991732513636353] [msgID=0] [reqSlices="[1]"]
[2022/05/05 06:34:05.504 +00:00] [DEBUG] [reduce.go:74] ["start marshal..."] [collectionID=432991732513636353] [msgID=0] [reqSlices="[1]"]
Assert "result_offset_end <= nq * topK" at /go/src/github.com/milvus-io/milvus/internal/core/src/segcore/reduce_c.cpp:185
=> illegal result_offset_end when GetSearchResultDataSlice, result_offset_end = 2, nq = 1, topk = 1Assert "result_offset_end <= nq * topK" at /go/src/github.com/milvus-io/milvus/internal/core/src/segcore/reduce_c.cpp:185
=> illegal result_offset_end when GetSearchResultDataSlice, result_offset_end = 2, nq = 1, topk = 1
free(): invalid pointer
free(): invalid pointer
SIGABRT: abort
PC=0x7fbc12c80e87 m=3 sigcode=18446744073709551610
goroutine 0 [idle]:
runtime: unknown pc 0x7fbc12c80e87
@sunby could you please take a look at this issue? /assign @sunby /unassign
/assign
may same as #16769
This issue is caused by appending wrong topk result when reduce search result. And this only happens when topK and nq are very small.
would fixed by #16815
verified and passed in version master-20220506-67ce206c
Is there an existing issue for this?
Environment
Current Behavior
Each time the code executes to the search step, querynode restarts, causing the search to fail. querynode's log is below. The strange thing is that there is no release operation performed in the code, but the error is reported in the log.
Expected Behavior
test case passed
Steps To Reproduce
Milvus Log
No response
Anything else?
logs: pod_log.zip