Skip to content

Latest commit

 

History

History
234 lines (146 loc) · 5.69 KB

CHANGELOG.md

File metadata and controls

234 lines (146 loc) · 5.69 KB
Raw

Change Log

All notable changes to this project will be documented in this file.

0.9.10 - 2015-03-06

  • Added a deb target to the Makefile using gox
  • Doc update

0.9.9 - 2015-03-06

Added

  • SHH_META: Off by default. This controls meta stats collection and reporting. Turn this on to get:

    • Librato stats (number of gauges, number of counters)
    • multi poller's stats
    • listen poller's meta stats

Removed

  • Removed lo from the default NIF_DEVICES
  • pid.last from the load poller. This is useless on a time series plot.

0.9.8 - 2015-01-05

Changed

  • Bugfix: ETS table memory to report in words not bytes.

0.9.7 - 2015-01-05

Changed

  • Bugfix: Used memory is a gauge, report it as such.

0.9.6 - 2015-01-05

Changed

  • Bugfix: Renamed LinearSliceContainsString to SliceContainsString, removing the old implementation, so that we only match on exact matches

0.9.5 - 2015-01-05

Added

  • Redis poller.

Changed

  • Update Go Version to 1.4.
  • Travis CI Updates

0.9.4 - 2014-12-16

Changed

  • Folsom Poller (folsom) now polls for ETS table size and memory.

0.9.3 - 2014-11-25

Changed

  • Folsom Poller (folsom) now uses native folsom_cowboy API calls to determine metric types.

0.9.2 - 2014-11-12

Added

  • Folsom Poller (folsom), which adds the ability to fetch metrics from a runinning Erlang system. See the docs for more info.

0.9.1 - 2014-11-11

  • Retry all low-level delivery errors

Changed (Breaking)

  • Fixed a typo in the per-processes rss metric. byts -> bytes

0.9.0 - 2014-11-07

Changed (Breaking)

  • MetricsNameNormalizer was changed so some metrics names may change.
    Old # -> _
        - -> _
    New # -> .
        _ -> -

0.8.9 - 2014-11-06

  • Actually catch io.EOFs so we can re-try them. Experience has shown they are most likely dropped connections.

0.8.8 - 2014-11-06

Added

  • Splunk Search Peers poller (splunksearchpeers), which adds the ability to monitor Splunk search clusters. See the docs for more info.

Changed (Breaking)

  • Removed LibratoNetworkTimeout, use NetworkTimeout (SHH_NETWORK_TIMEOUT) instead.
  • Merged listen poller metrics into 3 (all parse errors are treated the same) and added prefix meta to them all.

Changed (Non-breaking)

  • Refactored listen poller stats counters to avoid mutex
  • Removed a bunch of useless logging
  • handleListenConnection now part of Listen poller instead of bare.

0.8.7 - 2014-11-03

  • nagios2stats bugfix

0.8.6 - 2014-11-03

  • nagios3stats poller

0.8.5 - 2014-10-31

  • Fix reporting of per processes sys / user cpu

0.8.4 - 2014-10-30

  • Ignore processes w/o names, likely due to a process exiting between enumerating the directory entries and reading /proc//stat

0.8.3 - 2014-10-30

  • Generate additional process stats for processes that match SHH_PROCESSES_REGEX.
  • SHH_PROCESSES_REGEX: \A\z - Regex of process names to poll and extra additional measurements for
  • SHH_TICKS: 100 - cpu ticks per second. Default should be correct for most systems. see getconf CLK_TCK. Temporary until we use cgo to get it
  • SHH_PAGE_SIZE: 4096 - kernel page size. Default should be correct for most systems. See getconf PAGESIZE. Temporary until we use cgo to get it.

0.8.2 - 2014-10-28

  • SHH_LIBRATO_ROUND: true - round measurement times to nearest interval during submission

0.8.1 - 2014-10-24

Changed (Breaking)

  • Some disk poller stats were incorrectly being reported as Gauges.

0.8.0 - 2014-10-22

Added

  • SHH_DF_LOOP introduced to avoid loopback mounts from showing up in df poller (default false)
  • SHH_FULL introduced to add back full set of metrics from some pollers

Changed

  • Remove DEFAULT_SELF_POLLER_MODE in favor of SHH_FULL="self"
  • SHH_CPU_AGGR defaults to true, eliminating cpu metrics for all cores by default
  • SHH_DF_TYPES removes tmpfs by default
  • Default pollers to minimal set. Utilize SHH_FULL to get full set of metrics

0.7.0 - 2014-10-22

  • Update some defaults: SHH_INTERVAL=60s & SHH_LIBRATO_BATCH_TIMEOUT=10s

0.6.4 - 2014-10-22

  • Bugfix

0.6.3 - 2014-10-22

  • SHH_LIBRATO_BATCH_SIZE defaults to 500
  • SHH_LIBRATO_NETWORK_TIMEOUT defaults to 5s
  • SHH_LIBRATO_BATCH_TIMEOUT defaults to SHH_INTERVAL
  • Librato Outputter timeout doesn't start until there is a measurement

0.6.2 - 2014-10-22

  • Report to librato how many guages / counters are being reported in a batch

0.6.1 - 2014-10-22

  • $SHH_LISTEN_TIMEOUT now controls the timeout on the socket.
  • Librato outlet now reports a User-Agent header at the request of Librato.
  • Timeout errors to the librato api are now reported.
  • Better handling of ntpdate sub process error messages.

0.6.0 - 2014-10-20

  • shh-value cli tool for interacting with the unix socket.
  • Latest version of Go (1.3.3) used.
  • use github.com/heroku/slog for structured logging (extracted from shh originally).
  • LISTEN Poller documentation.
  • Improved Listen Poller with support for types and units
  • Use of Go's logger (over fmt.Println)

0.5.0 - 2014-10-09

Added

  • Units. Measurements now have units which the Librato outputter takes advantage of. No other outputters currently take advantage of this.

Changed (Breaking)

  • sockstat poller now uses lowercase protocol names in emitted metrics. Previously, it broke convention and used uppercase. (i.e. 'UDP' is now 'udp')
  • The percentage calculations for the "mem" and "df" pollers resulted in values between 0 and 1. CPU percentages are between 0-100. They now all use 0-100 instead of 0-1.

Prior Versions

Sadly, we didn't keep a proper changelog for previous versions. :(