Open honggyukim opened 1 year ago
I have seen many lost events during kernel tracing when the workload is heavy. I saw that the lost comes from the shmem buffer allocation failure at https://github.com/namhyung/uftrace/blob/v0.13.1/libmcount/record.c#L107-L116.
This makes the record make as LOST at https://github.com/namhyung/uftrace/blob/v0.13.1/libmcount/record.c#L165-L178.
LOST
We should investigate why shmem buffer allocation fails here.
This problem might be related to rlimit as @namhyung mentioned.
We can try to see if --num-thread or --rt-prio makes difference.
--num-thread
--rt-prio
I have seen many lost events during kernel tracing when the workload is heavy. I saw that the lost comes from the shmem buffer allocation failure at https://github.com/namhyung/uftrace/blob/v0.13.1/libmcount/record.c#L107-L116.
This makes the record make as
LOST
at https://github.com/namhyung/uftrace/blob/v0.13.1/libmcount/record.c#L165-L178.We should investigate why shmem buffer allocation fails here.