The Prometheus monitoring system and time series database.
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
 

7.3 KiB

0.11.1 / 2015-02-27

  • [BUGFIX] Make series maintenance complete again. (Ever since 0.9.0rc4, or commit 0851945, series would not be archived, chunk descriptors would not be evicted, and stale head chunks would never be closed. This happened due to accidental deletion of a line calling a (well tested :) function.
  • [BUGFIX] Do not double count head chunks read from checkpoint on startup. Also fix a related but less severe bug in counting chunk descriptors.
  • [BUGFIX] Check last time in head chunk for head chunk timeout, not first.
  • [CHANGE] Update vendoring due to vendoring changes in client_golang.
  • [CLEANUP] Code cleanups.
  • [ENHANCEMENT] Limit the number of 'dirty' series counted during checkpointing.

0.11.0 / 2015-02-23

  • [FEATURE] Introduce new metric type Histogram with server-side aggregation.
  • [FEATURE] Add offset operator.
  • [FEATURE] Add floor, ceil and round functions.
  • [CHANGE] Change instance identifiers to be host:port.
  • [CHANGE] Dependency management and vendoring changed/improved.
  • [CHANGE] Flag name changes to create consistency between various Prometheus binaries.
  • [CHANGE] Show unlimited number of metrics in autocomplete.
  • [CHANGE] Add query timeout.
  • [CHANGE] Remove labels on persist error counter.
  • [ENHANCEMENT] Various performance improvements for sample ingestion.
  • [ENHANCEMENT] Various Makefile improvements.
  • [ENHANCEMENT] Various console template improvements, including proof-of-concept for federation via console templates.
  • [ENHANCEMENT] Fix graph JS glitches and simplify graphing code.
  • [ENHANCEMENT] Dramatically decrease resources for file embedding.
  • [ENHANCEMENT] Crash recovery saves lost series data in 'orphaned' directory.
  • [BUGFIX] Fix aggregation grouping key calculation.
  • [BUGFIX] Fix Go download path for various architectures.
  • [BUGFIX] Fixed the link of the Travis build status image.
  • [BUGFIX] Fix Rickshaw/D3 version mismatch.
  • [CLEANUP] Various code cleanups.

0.10.0 / 2015-01-26

  • [CHANGE] More efficient JSON result format in query API. This requires up-to-date versions of PromDash and prometheus_cli, too.
  • [ENHANCEMENT] Excluded non-minified Bootstrap assets and the Bootstrap maps from embedding into the binary. Those files are only used for debugging, and then you can use -web.use-local-assets. By including fewer files, the RAM usage during compilation is much more manageable.
  • [ENHANCEMENT] Help link points to http://prometheus.github.io now.
  • [FEATURE] Consoles for haproxy and cloudwatch.
  • [BUGFIX] Several fixes to graphs in consoles.
  • [CLEANUP] Removed a file size check that did not check anything.

0.9.0 / 2015-01-23

  • [CHANGE] Reworked command line flags, now more consistent and taking into account needs of the new storage backend (see below).
  • [CHANGE] Metric names are dropped after certain transformations.
  • [CHANGE] Changed partitioning of summary metrics exported by Prometheus.
  • [CHANGE] Got rid of Gerrit as a review tool.
  • [CHANGE] 'Tabular' view now the default (rather than 'Graph') to avoid running very expensive queries accidentally.
  • [CHANGE] On-disk format for stored samples changed. For upgrading, you have to nuke your old files completely. See "Complete rewrite of the storage
  • [CHANGE] Removed 2nd argument from delta.
  • [FEATURE] Added a deriv function.
  • [FEATURE] Console templates.
  • [FEATURE] Added absent function.
  • [FEATURE] Allow omitting the metric name in queries.
  • [BUGFIX] Removed all known race conditions.
  • [BUGFIX] Metric mutations now handled correctly in all cases.
  • [ENHANCEMENT] Proper double-start protection.
  • [ENHANCEMENT] Complete rewrite of the storage layer. Benefits include:
    • Better query performance.
    • More samples in less RAM.
    • Better memory management.
    • Scales up to millions of time series and thousands of samples ingested per second.
    • Purging of obsolete samples much cleaner now, up to completely "forgetting" obsolete time series.
    • Proper instrumentation to diagnose the storage layer with... well... Prometheus.
    • Pure Go implementation, no need for cgo and shared C libraries anymore.
    • Better concurrency.
  • [ENHANCEMENT] Copy-on-write semantics in the AST layer.
  • [ENHANCEMENT] Switched from Go 1.3 to Go 1.4.
  • [ENHANCEMENT] Vendored external dependencies with godeps.
  • [ENHANCEMENT] Numerous Web UI improvements, moved to Bootstrap3 and Rickshaw 1.5.1.
  • [ENHANCEMENT] Improved Docker integration.
  • [ENHANCEMENT] Simplified the Makefile contraption.
  • [CLEANUP] Put meta-data files into proper shape (LICENSE, README.md etc.)
  • [CLEANUP] Removed all legitimate 'go vet' and 'golint' warnings.
  • [CLEANUP] Removed dead code.

0.8.0 / 2014-09-04

  • [ENHANCEMENT] Stagger scrapes to spread out load.
  • [BUGFIX] Correctly quote HTTP Accept header.

0.7.0 / 2014-08-06

  • [FEATURE] Added new functions: abs(), topk(), bottomk(), drop_common_labels().
  • [FEATURE] Let console templates get graph links from expressions.
  • [FEATURE] Allow console templates to dynamically include other templates.
  • [FEATURE] Template consoles now have access to their URL.
  • [BUGFIX] Fixed time() function to return evaluation time, not wallclock time.
  • [BUGFIX] Fixed HTTP connection leak when targets returned a non-200 status.
  • [BUGFIX] Fixed link to console templates in UI.
  • [PERFORMANCE] Removed extra memory copies while scraping targets.
  • [ENHANCEMENT] Switched from Go 1.2.1 to Go 1.3.
  • [ENHANCEMENT] Made metrics exported by Prometheus itself more consistent.
  • [ENHANCEMENT] Removed incremental backoffs for unhealthy targets.
  • [ENHANCEMENT] Dockerfile also builds Prometheus support tools now.

0.6.0 / 2014-06-30

  • [FEATURE] Added console and alert templates support, along with various template functions.
  • [PERFORMANCE] Much faster and more memory-efficient flushing to disk.
  • [ENHANCEMENT] Query results are now only logged when debugging.
  • [ENHANCEMENT] Upgraded to new Prometheus client library for exposing metrics.
  • [BUGFIX] Samples are now kept in memory until fully flushed to disk.
  • [BUGFIX] Non-200 target scrapes are now treated as an error.
  • [BUGFIX] Added installation step for missing dependency to Dockerfile.
  • [BUGFIX] Removed broken and unused "User Dashboard" link.

0.5.0 / 2014-05-28

  • [BUGFIX] Fixed next retrieval time display on status page.
  • [BUGFIX] Updated some variable references in tools subdir.
  • [FEATURE] Added support for scraping metrics via the new text format.
  • [PERFORMANCE] Improved label matcher performance.
  • [PERFORMANCE] Removed JSON indentation in query API, leading to smaller response sizes.
  • [ENHANCEMENT] Added internal check to verify temporal order of streams.
  • [ENHANCEMENT] Some internal refactorings.

0.4.0 / 2014-04-17

  • [FEATURE] Vectors and scalars may now be reversed in binary operations (<scalar> <binop> <vector>).
  • [FEATURE] It's possible to shutdown Prometheus via a /-/quit web endpoint now.
  • [BUGFIX] Fix for a deadlock race condition in the memory storage.
  • [BUGFIX] Mac OS X build fixed.
  • [BUGFIX] Built from Go 1.2.1, which has internal fixes to race conditions in garbage collection handling.
  • [ENHANCEMENT] Internal storage interface refactoring that allows building e.g. the rule_checker tool without LevelDB dynamic library dependencies.
  • [ENHANCEMENT] Cleanups around shutdown handling.
  • [PERFORMANCE] Preparations for better memory reuse during marshalling / unmarshalling.