Matt Sergeant wrote:
To be honest John I missed all the adaptive logging stuff due to working
on high_perf. I'm confused as to why you buffer up the log lines like
this. Can you explain?
Sure. If you want to log at one level for success and once level for
failure, you need to actually wait until you know whether a given
message will fall into which bucket. The only way that made sense was
to buffer the lines up and emit only at the point where it is possible
to know.
I want to have two log files:
1) a slowly updated one which only logs success (so when the boss calls
and wants to make sure some message went out last week, I can say that
it did);
2) a faster updated log of failures (which can get rolled more quickly
because I usually find out about problems sooner).
As I said, I think I can get #2 by just emitting everything below max
and get #1 by re-emitting the success lines (with something I can filter
on in multilog) when the message is actually delivered.
John
