Christos Kalkanis
Christos Kalkanis
> > * Network traffic overhead for self-contained messages is estimated to be about 3.5x, after compression (and neglecting any cost of compression and decompression). > > * The uncompressed...
Adding a comment to keep this PR from being closed.
Could someone with rights (@mx-psi @AlexanderWert maybe?) to do so reopen this pull request?
> > We can increase the limit further (e.g. 256MB?) to support binaries such as these. > > This will obviously temporarily bloat the profilers memory usage, and potentially cause...
> In this code path we use an unbounded map - maybe this is the reason for the increased memory usage: I discussed this a little with @fabled today, warrants...
@lgwozdecki Did you get a chance to test?
Closing this issue as, for now, we don't plan to add filtering inside the eBPF profiler. As @florianl mentioned, downstream consumers can still filter the data inside the OTel collector...
> Should default sample type go to the scope message? That would be less flexible as it'd force us to have separate instrumentation scopes (`ScopeProfiles`) for different sample types. With...
> > Should default sample type go to the scope message? > > I mostly want to make sure there is some way to capture the default sample type. The...
> The question is how such a profile with four sample types and one of them being the preferred default as chosen by the profile producer should be encoded in...