Add the option to randomly assign UIDs to metrics to improve distribution across
HBase region servers.
Introduce salting of data to improve distribution of high cardinality regions
across region servers.
Introduce query stats for tracking various timings related to TSD queries.
Add more stats endpoints including /threads, /jvm and /region_clients
Allow for deleting UID mappings via CLI or the API
Name the various threads for easier debugging, particularly for distinguishing
between TSD and AsyncHBase threads.
Allow for pre-fetching all of the meta information for the tables to improve
performance.
Update to the latest AsyncHBase with support for secure HBase clusters and RPC
timeouts.
Allow for overriding metric and tag widths via the config file. (Be careful!)
URLs from the API are now relative instead of absolute, allowing for easier reverse
proxy use.
Allow for percent deviation in the Nagios check
Let queries skip over unknown tag values that may not exist yet (via config)
Add various query filters such as case (in)sensitive pipes, wildcards and pipes
over tag values. Filters do not work over metrics at this time.
Add support for writing data points using Appends in HBase as a way of writing
compacted data without having to read and re-write at the top of each hour.
Introduce an option to emit NaNs or Nulls in the JSON output when downsampling and
a bucket is missing values.
Introduce query time flags to show the original query along with some timing stats
in the response.
Introduce a storage exception handler plugin that will allow users to spool or
requeue data points that fail writes to HBase due to various issues.
Rework the HTTP pipeline to support plugins with RPC implementations.
Allow for some style options in the Gnuplot graphs.
Allow for timing out long running HTTP queries.
Text importer will now log and continue bad rows instead of failing.
New percentile and count aggregators.
Add the /api/annotations endpoint to fetch multiple annotations in one call.
Add a class to support improved bulk imports by batching requests in memory for a
full hour before writing.
Allow overriding the metric and tag UID widths via config file instead of
having to modify the source code.
Rework the QueryStats output to be a bit more useful and add timings from the
various scanners and query components.
Modify the UI to allow for group by or aggregate per tag (use the new query feature)
Rework the UI skin with the new TSDB logo and color scheme.
Add the QueryLog config to logback.xml so users can optionally enable logging of
all queries along with their stats.
v2.1.4
8 years ago
Version 2.1.4 (2016-02-14)
Bug Fixes:
Fix the meta table where the UID/TSMeta APIs were not sorting tags properly
prior to creating the row key, thus allowing for duplicates if the caller changed
the order of tags.
Fix a situation where meta sync could hang forever if a routine threw an exception.
Fix an NPE thrown when accessing the /logs endpoint if the Cyclic appender is not
enabled in the logback config.
Remove an overly chatty log line in TSMeta on new time series creation.
v2.2.0RC3
8 years ago
Version 2.2.0 RC3 (2015-11-11)
Bug Fixes:
Fix build issues where the static files were not copied into the proper location.
v2.1.3
8 years ago
Version 2.1.3 (2015-11-11)
Bug Fixes:
Fix build issues where the static files were not copied into the proper location.
v2.2.0RC2
8 years ago
Version 2.2.0 RC2 (2015-11-09)
Noteworthy Changes:
Allow overriding the metric and tag UID widths via config file instead of
having to modify the source code.
Bug Fixes:
OOM handling script now handles multiple TSDs installed on the same host.
Fix a bug where queries never return if an exception is thrown from the
storage layer.
Fix random metric UID assignment in the CLI tool.
Fix for meta data sync when salting is enabled.
v2.1.2
8 years ago
Bug Fixes:
Fix the built-in UI to handle query parameter parsing properly (found when Firefox
changed their URI behavior)
Fix comments about the Zookeeper quorum setting in various config files.
Fix quoting in the Makefile when installing.
Make sure builds write files in the proper location on FreeBSD.
v2.2.0RC1
8 years ago
Version 2.2.0 RC1 (2015-09-12)
Noteworthy Changes:
Add the option to randomly assign UIDs to metrics to improve distribution across
HBase region servers.
Introduce salting of data to improve distribution of high cardinality regions
across region servers.
Introduce query stats for tracking various timings related to TSD queries.
Add more stats endpoints including /threads, /jvm and /region_clients
Allow for deleting UID mappings via CLI or the API
Name the various threads for easier debugging, particularly for distinguishing
between TSD and AsyncHBase threads.
Allow for pre-fetching all of the meta information for the tables to improve
performance.
Update to the latest AsyncHBase with support for secure HBase clusters and RPC
timeouts.
Allow for overriding metric and tag widths via the config file. (Be careful!)
URLs from the API are now relative instead of absolute, allowing for easier reverse
proxy use.
Allow for percent deviation in the Nagios check
Let queries skip over unknown tag values that may not exist yet (via config)
Add various query filters such as case (in)sensitive pipes, wildcards and pipes
over tag values. Filters do not work over metrics at this time.
Add support for writing data points using Appends in HBase as a way of writing
compacted data without having to read and re-write at the top of each hour.
Introduce an option to emit NaNs or Nulls in the JSON output when downsampling and
a bucket is missing values.
Introduce query time flags to show the original query along with some timing stats
in the response.
Introduce a storage exception handler plugin that will allow users to spool or
requeue data points that fail writes to HBase due to various issues.
Rework the HTTP pipeline to support plugins with RPC implementations.
Allow for some style options in the Gnuplot graphs.
Allow for timing out long running HTTP queries.
Text importer will now log and continue bad rows instead of failing.
New percentile and count aggregators.
Add the /api/annotations endpoint to fetch multiple annotations in one call.
Add a class to support improved bulk imports by batching requests in memory for a
full hour before writing.
Bug Fixes:
Modify the .rpm build to allow dashes in the name.
Allow the Nagios check script to handle 0 values properly in checks.
Fix FSCK where floating point values were not processed correctly (#430)
Fix missing information from the /appi/uid/tsmeta calls (#498)
Fix more issues with the FSCK around deleting columns that were in the list (#436)
Avoid OOM issues over Telnet when the sending client isn't reading errors off it's
socket fast enough by blocking writes.
2.1.1
8 years ago
Version 2.1.1 (2015-09-12)
Bug Fixes:
Relax the pgrep regex to correctly find and kill the java process in the RPM init.d
script.
Improve query performance slightly when aggregating multiple series.
Fix the /api/search/lookup API call to properly handle the limit parameter.
Fix the /api/query/last endpoint to properly handle missing tsdb-meta tables.
v2.1.0
9 years ago
Version 2.1.0 (2015-05-06)
Bug Fixes:
FSCK was not handling compacted and floating point duplicates properly. Now they
are merged correctly.
TSMeta data updates were not loading the latest data from storage on response
The config class will now trim spaces from booleans and integers
On shutdown, the idle state handler could prevent the TSD from shutting down
gracefully. A new thread factory sets that thread as a daemon thread.
TSMeta objects were not generated if multiple writes for the same data point arrived
in succession due to buffering atomic increments. Increments are no longer buffered.
Updated paths to the deprecated Google Code repo for dependencies.
v2.1.0RC2
9 years ago
Version 2.1.0 RC1 (2015-04-04)
Noteworthy Changes:
Handle idle connections in Netty by closing them after some period of inactivity
Support compressed HTTP responses
Bug Fixes:
Various RPM script and package fixes
Properly handle deletions of the cache directory while running
Queries for non-extant metrics now return a 400 error
Fix an issue with the meta sync utility where entries were not created if the counter existed
Report stats properly when the UID is greater than 3 bytes wide
Fix UI hangups when incorrect tags are supplied
Log illegal put exceptions at the debug level
Fix global annotation retrieval where some entries were missing
Fix unit tests that were not properly mocking out clients or threads and causing JVM OOMs
Fix accounting bugs in FSCK
Sort aggregators in the UI
Properly throw an exception if the user supplies an empty tag on storage or retreival