This is too late if the process already exited. With the default profiling duration of 60s, this is common.
For process profiles (_profile_process), we already get the application metadata & appid when the profile begins, so we always get them.
For perf profiles, we don't have the PIDs that were profiled until we run perf script. We can just collect container names for ALL live processes, but then we have a race with PID reuse, not sure how likely is that for a 60s profiling session. If perf could output a stronger identifier, e.g PID+start-time, it'd help.
We enrich container names when merging profilers: https://github.com/Granulate/gprofiler/blob/c2899341813c5db57b8401356c02c74cc6ec75d5/gprofiler/merge.py#L318-L319
This is too late if the process already exited. With the default profiling duration of 60s, this is common.
_profile_process), we already get the application metadata & appid when the profile begins, so we always get them.perf script. We can just collect container names for ALL live processes, but then we have a race with PID reuse, not sure how likely is that for a 60s profiling session. Ifperfcould output a stronger identifier, e.g PID+start-time, it'd help.