[erlang-questions] Should we deprecate or modernize log_mf_h and rb?

Vance Shipley vances@REDACTED
Sat Apr 2 09:05:32 CEST 2016


On Apr 1, 2016 12:44 PM, "Siri Hansen" <erlangsiri@REDACTED> wrote:
> Are you using log_mf_h (e.g. by setting environment variables 'error_logger_mf_*' in sasl) and rb?

Yes, this is part of my default configuration for embedded systems.

> If so, why - is it the log rotation or the binary logging (or both) that you are really after?

Both.

> Have you considered alternative tools?

I've considered application specific versions of this scheme.

On Sat, Apr 2, 2016 at 1:23 AM, Robert Raschke <rtrlists@REDACTED> wrote:
> Unfortunately, straight logged binary data and formatting it at read time appears to be majorly out of fashion :-(

Stupidly so IMHO.  I often see systems with a mission of moving
massive amounts of message traffic logging data orders of magnitude
larger than the size of the actual message.  Doing so naively results
in the servers spending much more time logging than doing the work
they're intended to do.  Logging is important and at this rate it
costs real money to accomplish so it deserves to be engineered
sensibly.  My rational is that where a gen_fsm is spawned for each
transaction it should simply do it's work and before terminating send
it's State variable term, usually a record, directly to a log handler
(e.g. disk_log).  To have the worker for traffic doing string
processing for an human to possibly read a long time later makes no
sense.  Log binary data to disk and format just the data requested, at
request time, formatted for the requestor.

-- 
     -Vance



More information about the erlang-questions mailing list