- Fix NFS export id caching
- The code checked for an entry in the cache, but would never find one since it didn't actually update the cache after a lookup
- fix nil dereference in v0.26
Warning
The Prometheus collector crashes in this release; use v0.27 instead which contains a fix for this issue
- Fix handling of pinned workloads for Prometheus
- Pinned workloads are a special workload type
- The Prometheus code did not understand this workload type and so it dropped the stats with an error
- The fix now retains pinned stats and adds a new tag to all regular stats named "pinned" value "true" or "false"
- Added support for InfluxDBv2 as a backend while retaining support for InfluxDBv1
- Fixed build issues with InfluxDBv1 libraries
- Fixed missing config compatibility in v0.24
- Added the ability to preserve the case of cluster names instead of normalizing to lower case
- Added configurable retry interval and count to calls to write to the configured backend database
Important
The config rewrite is a breaking change which will require a manual update of the configuration file post-upgrade.
- Added secret support and log to stdout support
- Added secret support for sensitive information such as cluster passwords
- Added logfile configuration to the config file
- Logging to stdout is enabled via the log_to_stdout parameter
- Logging to file can be configured in the config file and overridden at the command line
- Added basic http landing page with description and link to "/metrics" endpoint
- Added support for tls-encrypted connections to the Prometheus endpoint
Important
The config rewrite is a breaking change which will require a manual update of the configuration file post-upgrade.
- Major config rewrite
- This config change means older config files must be updated to be compatible
- Renamed back end names (removed "_plugin")
- Removed hacky "processor_args" inherited from the Python collector
- Added config stanzas for InfluxDB and Prometheus
- Added missing "lookup_export_ids" setting
- Added config version checking
- Upcoming changes will break the config file format.
- Added a version check to avoid unexplained breakage.
- Updated all dependencies to latest versions
- Restored missing "isilon" namespace from stats
- In the switch to using our own Collector interface rather than the pre-canned gauges, the metric names lost their namespace prefix
- Reworked Prometheus support
- The initial Prometheus support had a major issue where stats that had been collected at least once, but which did not appear in the current collection cycle were still exposed via the
/metricsendpoint - Completely rewrote the code. Advantages of the rewritten code include
- the collector now correctly exports the original metric timestamp for each metric, and
- the collector now expires metrics based on the metadata that defines how long they are value so stale metrics correctly disappear from the
/metricsendpoint
- The initial Prometheus support had a major issue where stats that had been collected at least once, but which did not appear in the current collection cycle were still exposed via the
- Added support for NFS export id to path lookup
- Optionally enabled via the config file.
- If enabled, it is necessary to grant readonly NFS API access to theconfigured stats user.
- Added support for hardcoding the HTTP SD listen addr
- Fixed HTTP SD output
- Added support for Prometheus HTTP SD discovery
Fixed http routing with multple servers
- Fixed prometheus plugin argument handling
- example configuration filename changed to match tool name
- Fixed cluster name breakage introduced by prom work
Version bump to sync with the current version of gostats
- Add Prometheus support
- Add support for authenticated InfluxDB connection