Out of interest, what log-level is recommended for live, high-throughput (50 SMS/sec+) SMPP connections?
Running SMPP connections with log-level=0 gives me a very good level of comfort as the full PDUs are in the log which makes it very easy to trace down issues (and send excerpts of to telcos/providers as evidence when things go wrong). This obviously generates copious amounts of logs, but with help from Linux's logrotate and a good backup/purge plan things aren't too bad. My main concern is how much throughput I'm losing with this log level. Our limited testing so far has shown that there's not much impact (maybe 10%) and especially as our SMPP peers are half-a-world away, the latency on the link is likely to have a much higher impact on throughput than log writing. I'm interested in hearing what log-level others run in production. Any stories? Obviously one option is to set log-level=0 when things need to be traced, but in the world of SMS where many issues are time-sensitive and messages themselves are sensitive and can't just be re-sent for "testing" purposes, it's much easier to have a log of what happened than trying to duplicate the issue, hence my preference for massive logs. The other interesting part is that there's a world of difference between log-level=0 and log-level=1. I would have expected to get basic information with a log-level=1 trace (e.g. a log of each PDU sent/received without a full dump of the contents), but that's not the case. With log-level=1 unless something goes quite wrong, I get empty logs other than startup/shutdown notifications. Anyone interested in a more traceable but less verbose log level? Cheers, Toby.
