Releases: OpenTSDB/asynchbase
Releases · OpenTSDB/asynchbase
v1.8.2
- Version 1.8.2 (2018-05-19)
This is a bug fix release with one beta feature.
- Fix a possible issue with the security config when spaces are present in the
authentication type. - Add beta support for split-meta from HBASE-11288 for large, multi-tenant
clusters. The table names may change in the future. - Auto-detect HBase 2.0 by catching an exception thrown when getClosestRowOrBefore
is called. HBASE-20237. - Log the version setting, min and max timestamps in Scanners.
- Add CallQueueTooBig retry handling in the RegionClient without NSREing.
v1.8.1
- Version 1.8.1 (2018-03-18)
This is a bug fix release.
Noteworthy bug Fixes:
- Revert locking in an RPC to synchronizing on the key instead of the entire
object as well as setting the suspend probe flag to volatile. - Fix GetRequests where the family was masked and immutable after being set.
- Fix setting the scanner reverse flag in the setReversed() setter.
Previously when called it would only set the flag totrue
. - Add RegionTooBusyException to allow for retries when newer versions of HBase
send this exception. - Fix processing of encoded (private) RPCs and catch + log exceptions during
SASL negotiation. - Add a configuration flag 'hbase.meta.scan' to switch meta lookups to
use a reverse scan when finding tables for compatibility with HBase 2.x
where the old method of getting the closest row was disabled. - Fix an issue with AssertionErrors being thrown when a multi-action RPC
had an exception of a type that didn't implement the make() method. - Fix an issue with multi-actions sent to HBase 1.31 an later where the
request was sorted by row key so that responses no longer matched the
request order. Now we sort by row key as well.
v1.8.0
- Version 1.8.0 (2017-06-09)
This is feature release with new APIs and behaviors.
New Features:
- Support HBase 1.3.0 and later where scanner behavior changed.
- Add optional support for buffered AtomicIncrement calls to bypass the WAL
for quicker writes. - Add an API for writing AtomicIncrements over multiple columns in a single
RPC. - Add the MultipleColumnPrefixFilter.
- Add a "multi-get" API that allows for batching GetRequests and sending them
in a single request. Useful if the client wants to fetch a number of
disparate rows when scanning would be less efficient. - Support for reverse scanning
v1.7.3
- Version 1.7.3 (2017-06-09)
This is a bug fix release.
Noteworthy bug Fixes:
- Properly handle IPV6 addresses in the RegionClientPipeline.
- Avoid a deadlock when calling the RegionClient.toString() method.
- Avoid a synchronization deadlock handling NSRE probes.
v1.7.2
- Version 1.7.2 (2016-09-17)
This is a bug fix release.
Noteworthy bug Fixes:
- Fix a race condition where messages in the NSRE queue could be stuck and
never retried properly. This will help region server crash recovery. - Fix a bug in batched RPCs where, on exception, the batch RPC itself was
returned in the exception instead of the individual RPCs. - Changed watermark variables to integers instead of shorts to avoid
overflows when set to values over 32K. - Added retries of CallQueueTooBigExceptions. These still need to avoid
flushing the region cache but at least they'll be retried. - Fix calls to scanner.setFamilies() where NPEs were being thrown.
- Retry RPCs on RegionServerAbortedException, ServerNotRunningYetExceptions
and RegionServerStoppedExceptions. - Fix CompareAndSetRequests so they serialize timestamps as well as the data
to target specific versions. - Update the protobuf compiler plugin for Maven.