Reduce tcp_syslog_send_timeout_secs to 4s from 30.
We no-longer need a logplex_token process.
Drains are now responsible for their own lookups.
Speed up populat_token_drain_data.
No longer resolves hostnames here.
Replace is_record() guard with pattern.
Really enable SMP this time.
Correctly record stats for messages in cb scenario.
erts: enable SMP for logplex erlang nodes.
Adds the v34-v35 upgrade scripts.
Correctly process messages for callbacks in tcp_proxy.
Define a default proxy message cb of logplex_worker.
Allow tcp_proxy to handle messages with a callback.
Export logplex worker state_init and handle_message.
Move lp_worker message handling to separate function.
Move logplex_worker state init to separate function.
Merge branch 'v34'
Add is_process alive check to ChildPids - generally avoids transient …
Revert "Count logs for unknown tokens."
This reverts commit 5d1050b.
Count logs for unknown tokens.
Adds support for class/key/value stats
Fix bogus stats entries.
Ensure we pass correct drain_ids.
Merge branch 'v34' of github.com:heroku/logplex into v34
Alter the code_change function to work correctly.
Fixes a crash where a v34 state would get upgraded twice incorrectly.
Update nsync_callback too.
Fix drain lookup.
Reduce nsync verbosity.
Fix float printing in duration data.
Add version tags to upgrade messages.
Adds a v33->v34 upgrade set of shell scripts.
Report duration times from tcpsyslog_drain.
Adds a connect_time (establishment time) to tcpsyslog.
Connection duration reporting in tcp_proxy:
* Adds connect_time to state
* Adds v33-34 upgrade path
* Adds duration reporting to tcp closed and error states.
Remove the v32-33 code_change path.
Adds the v32->v33 upgrade shellcode.