flexible tracing as alternative to logging?

[Loïc Hoguin]

Tracing based tools would be interesting to have. The problem currently 
is that multiple tools can easily conflict with each other, and conflict 
with any trace debugging session you'd like to do in the shell. So I 
would not recommended it.

On 29/05/2020 16:56, Vans S wrote:
> Has anyone ever ran into implementing something along the lines of using 
> tracing to log.
> For example say we spawn 1000 processes, now each of these spawn a 
> linked process
> called trace_logger, which places traces on the parent based on a match 
> spec.
> Each of the 1000 procs has their own child proc link trace_logger (not a 
> global tracer, to prevent a bottleneck).
> 
> Then this process can now implement its own filters based on the traces 
> coming in, like say if a certain function gets
> called with the wrong argument, log a logger error.
> 
> Would tracing be performant enough to achieve this, even if the cost 
> penalty is significant it might be OK and worth it.
> 
> Thanks

-- 
Loïc Hoguin
https://ninenines.eu