CHANGELOG.md

v0.16.0

• fix: add error to reset command result so it is honored
• fix: remove deadlock in setReverseConfig
• upd: revert max reverse conn retries back to infinite (-1)
• upd: switch server request logging to debug level
• fix: handle enable new metrics when there's an error with the state directory. Disable new metric enabling and turn off check management, but allow check to be created.
• upd: export plugin/builtin stat times as strings for json

v0.15.1

• fix: lock contention in check refresh
• upd: optimize lastMetrics (val->ptr)
• upd: additional debugging during run, flush, package
• upd: concurrent flushes (builtins, plugins, receiver, statsd, prom receiver)
• fix: var name typo
• upd: lastMetrics handling

v0.15.0

• mrg: tracking upstream
• add: check bundle id to all reverse log messages
• upd: fatal when errors starting main listening servers (e.g. address already in use)
• upd: add more debug lines around refreshing check configuration
• fix: lock contention when refreshing check configuration for reverse AND enable new metrics is active
• add: --reverse-max-conn-retry command line option
• wip: gate max requests from broker before forced reset
• wip: refactor of inventory endpoint
• upd: cosi v2 prep

v0.14.1

• upd: io latency plugin for linux - add additional signals to handler and cleanup in tracing start

v0.14.0

• add: io latency plugin for linux

v0.13.4

• upd: upstream deps (cgm)

v0.13.3

• upd: upstream deps

v0.13.2

• fix: linxu/procfs/diskstats collector. guard against malformed/blank lines
• doc: miscellaneous documentation updates

v0.13.1

• fix: remove version line from --show-config output
• new: api package

v0.13.0

• upd: refactor enable new metric logic to work better for new check use case
• upd: refactor reverse connection and command processing

v0.12.0

• add: -E as a short option for --check-enable-new-metrics
• add: --check-metric-state-dir configuration option
• doc: stream tag support for JSON plugin output and JSON receiver
• add: stream tag support to parsers for JSON plugin output and JSON receiver /write/... endpoint
• doc: stream tag support to exec plugin output
• add: stream tag support to exec plugin output parser
• doc: stream tag support for StatsD receiver
• add: stream tag support to StatsD receiver metric parser
• upd: fix StatsD metric syntax for rates, require |@
• upd: optimize label handling in prometheus receiver/collector
• add: tags package, centralize handling of tag list to stream tag spec
• upd: allow underscore in plugin name with /run/... endpoint
• add: error message when both --check-create and --check-id options are specified
• upd: build constraint to go1.10

v0.11.0

• upd: remove deprecated options/defaults
• doc: clarify comment in frame handling
• upd: honor RESETs from broker on reverse
• upd: use stream tag syntax by default for supported sources
• add: warn logging on duplicate metric names
• add: debug log summary of lines processed, metrics found, errors, and duplicates found while processing exec plugin output
• add: --debug-dump-metrics option to dump json sent to broker (option argument is a directory name where the user running circonus-agentd has write permissions. output file name format metrics_ccyymmdd_hhmmss.json)
• add: more tests to check package (state)
• add: more tests to check package (metrics)
• upd: switch from using map[string]interface{} to only using cgm.Metrics for metrics
• fix: issue with incrementing offset by bytes sent rather than payload length
• upd: check.New func signature in reverse tests
• upd: dependencies
• upd: remove httpgzip constraint
• add: tests for ParseListen
• doc: comment ParseListen

v0.10.0

• fix: revert to check bundle updates vs check bundle metrics endpoint until histogram issue resolved
• add: state directory requirement for enable new metrics feature
• add: ability to automatically enable new metrics
• add: ability to have agent create a new check when not using reverse
• upd: reorganize check handling into dedicated package
• upd: ⚠️ BREAKING changes to configuration options check configuration options changed/reorganized
• add: example systemd service configuration to release tarball
• add: --no-gzip to force disable gzip'd responses
• add: Transfer-Encoding: identity to non-chunked responses
• add: fix chunked gzip responses, broker does not understand them

v0.9.2

• upd: add circleci for integration testing
• upd: add linux arm release FOR TESTING (e.g. raspbian on pi3)
• upd: replace deprecated tr.CancelRequest w/context
• upd: defer context cancel() call
• fix: several test t.Fatalf calls
• upd: simplify (gofmt -s)
• fix: var shadow instances
• fix: typos in messages and comments
• doc: fix configuration link in main README.md
• doc: clarify configuration file quick start instructions in etc/README.md
• doc: fix typo (cirocnus) etc/README.md
• upd: dependencies
• upd: import path for httpgzip

v0.9.1

• upd: constrain IDs provided in /run and /write URLs to [a-zA-Z0-9-]; ensure clean metric name prefixes
• fix: make plugins optional, do not fatal error if no plugins are found

v0.9.0

• add: initial mvp prometheus collector (pull prometheus text formatted metrics from an endpoint)
• add: initial mvp prometheus receiver (accept pushed prometheus text formatted metrics)

v0.8.0

• fix: socket server test to fail correctly when on a linux vagrant mounted fs
• fix: various tests for cross platform differences
• fix: several tests to accommodate differences in os error messages across platforms
• add: bytes read/written to procfs.diskstats (eliminate need for nad:linux/disk.sh)
• fix: normalize procfs.cpu clockHZ
• fix: double backtick procfs.if tcp metrics
• upd: expose all procfs.vm meminfo metrics as raw names
• add: linux default collectors ['cpu', 'diskstats', 'if', 'loadavg', 'vm']
• add: procfs.loadavg collect (nad:common/loadavg.elf)
• add: procfs.vm collector (nad:linux/vm.sh)
• add: procfs.if collector (nad:linux/if.sh)
• add: procfs.diskstats collector (nad:linux/diskstats.sh)
• add: procfs.cpu collector (nad:linux/cpu.sh)
• doc: etc/README.md and plugins/README.md to releases
• doc: reorganize documentation
• doc: add plugins/README.md with documentation about plugins
• doc: add etc/README.md with documentation specifically about configuring the Circonus agent and builtin collectors
• upd: vendor deps

v0.7.0

• add: builtin collector framework
• new: configuration option --collectors (for platforms which have builtin collectors - windows only in this release)
• add: wmi builtin collectors for windows
  • available WMI collectors: cache, disk, interface, ip, memory, object, paging_file, processes, processor, tcp, udp
  • default WMI collectors enabled ['cache', 'disk', 'ip', 'interface', 'memory', 'object', 'paging_file' 'processor', 'tcp', 'udp']
• new: collectors take precedence over plugins (e.g. collector named cpu would prevent plugin named cpu from running)
• upd: plugin directory is now optional - valid use case to run w/o plugins - e.g. only builtins, statsd, receiver or a combination of the three
• upd: select fastest broker rather than picking randomly from list of all available brokers. If multiple brokers are equally fast, fallback to picking randomly, from the list of fastest brokers.
• fix: use value of --reverse-target (if specified) in --reverse-create-check-title (if not specified)

v0.6.0

• exit agent is issue creating/starting any server (http, ssl, sock)
• config file setting renamed plugin-dir -> plugin_dir to match other settings
• add unix socket listener support (for /write endpoint only)
  • command line option, one or more, -L </path/to/socket_file> or --listen-socket=</path/to/socket_file>
  • config file listen_socket (array of strings)
  • handle encoded histograms (e.g. cgm sending to agent /write endpoint)
• add ttl capability to plugins; parsed from plugin name (e.g. test_ttl5m.sh run once every five minutes) valid units ms, s, m, h.
• allow multiple listen ip:port specs to be used (e.g. -l 127.0.0.1:2609 -l 192.168.1.2:2630 ...)
• migrate configuration validation to (server|statsd|plugins).New functions
• docs: add new --plugin-ttl-units default s[econds]
• docs: add new -L, --listen-socket
• docs: update --listen allow more than one
• docs: update --show-config now requires a format to output (json|toml|yaml)

Socket example:

# start agent with the additional setting
$ /opt/circonus/agent/sbin/circonus-agentd ... --listen-socket=/tmp/test.sock

$ curl --unix-socket /tmp/test.sock \
    -H 'Content-Type: application/json' \
    -d '{"test":{"_type":"i","_value":1}}' \
    http:/circonus-agent/write/socktest

# resulting metric: socktest`test numeric 1

Example configuring CGM to use an agent socket named /tmp/test.sock:

cmc := &cgm.Config{}

cmc.CheckManager.Check.SubmissionURL = "http+unix:///tmp/test.sock/write/prefix_id"
// prefix_id will be the first part of the metric names

cmc.CheckManager.API.TokenKey = ""
// disable check management and interactions with API

client, err := cgm.New(cmc)
if err != nil {
    panic(err)
}

client.Increment("metric_name")

// resulting in metric: prefix_id`metric_name numeric 1

v0.5.0

• standardize on cgm.Metric(s) structs for all metrics
• strict parsing of JSON sent to receiver /write endpoint
• add /prom endpoint (candidate poc)
• improve handling of invalid sized reads
• group plugin stats
• update cgm version
• common appstats package

v0.4.0

• switch --show-config to take a format argument (json|toml|yaml)
• add more test coverage
• reorganize README
• update dependencies

v0.3.0

• add running config settings to app stats
• add env vars for options to help output
• switch env var prefix from NAD_ to CA_

v0.2.0

• add ability to create a [reverse] check - if a check bundle id is not provided for reverse, the agent will search for a suitable check bundle for the host. Previously, if a check bundle could not be found the agent would exit. Now, when --reverse-create-check is supplied, the agent has the ability to create a check, rather than exit.
• expose basic app stats endpoint /stats

v0.1.2

• fix statsd packet channel (broken in v0.1.1)
• update readme with current instructions

v0.1.1

• merge structs
• eliminate race condition

v0.1.0

• add freebsd and solaris builds for testing
• add more test coverage throughout
• switch to tomb instead of contexts
• refactor code throughout
• add build constraints to control target specific signal handling in agent package
• fix race condition w/inventory handler
• reset connection attempts after successful send/receive (catch connection drops)
• randomize connection retry attempt delays (not all agents retrying on same schedule)

v0.0.3

• integrate context
• cleaner shutdown handling

v0.0.2

• move circonus-agentd binary to sbin/circonus-agentd
• refactor (plugins, server, reverse, statsd)
• add agent package

v0.0.1

• Initial development working release