Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Threaded View
HBase >> mail # dev >> [ANNOUNCE] The 1st hbase 0.94.10 release candidate is available for download


Copy link to this message
-
Re: [ANNOUNCE] The 1st hbase 0.94.10 release candidate is available for download
Here we go ;)

Downloaded, validated the signatures and uncompressed the files. => Passed.
On the README.txt file, checked link for 6. => 404 page not found. since
current directory doesn't exist, other links will fail too.
Checked the CHANGE.txt file => Seems to be ok.
Randomly checked the documentation => docs/bulk-loads.html point to an
external page. Should it not be to a local one?
Ran all tests => Passed.
Started it, looked at the logs, the UI, all is fine.
Ran PE, results similar between 0.94.9 and 0.94.10 (Numbers below are ms
spent to run the test, average on 10 tests.)
||Test||0.94.9||0.94.10||
|org.apache.hadoop.hbase.PerformanceEvaluation$FilteredScanTest|2172297.6|2141379.4|
|org.apache.hadoop.hbase.PerformanceEvaluation$RandomReadTest|2574442.3|2533646.4|
|org.apache.hadoop.hbase.PerformanceEvaluation$RandomScanWithRange100Test|19098.3|19205|
|org.apache.hadoop.hbase.PerformanceEvaluation$RandomSeekScanTest|154499.7|153022.8|
|org.apache.hadoop.hbase.PerformanceEvaluation$RandomWriteTest|84760.2|85176.2|
|org.apache.hadoop.hbase.PerformanceEvaluation$SequentialWriteTest|43636.175|44185.7375|

Did rolling restart on a 8 node cluster, restarted the master => Ok.
Checked the logs and UI => Ok.
Stopped master, configured balancer to default, merged a 71 region table
into one single region => Ok.
Major compacted the table, got it automatically splited to 2 regions,
manually split it to 32 regions, waited for balancer => Ok.
Stopped master again, reconfigured custom balancer, restarted, rebalanced
=> Ok.
>From shell: Create table, scan, balance, drop => Ok.

IntegrationTestLoadAndVerify still give strange results =>
REFERENCES_CHECKED=9855348 (Should be 10000000)
Ran IntegrationTestBigLinkedList => Perfect.

So a big overall +1 for me ;)

Deployed in my "prod" cluster. Thanks all for this release.

2013/7/19 lars hofhansl <[EMAIL PROTECTED]>

> The 1st 0.94.10 RC is available for download at
> http://people.apache.org/~larsh/hbase-0.94.10-rc0/
> Signed with my code signing key: C7CFE328
>
> Like the previous point releases, 0.94.10 is a primarily a bug fix
> release. This RC has 35 issues resolved against it:
>     [HBASE-8774] - Add BatchSize and Filter to Thrift2
>     [HBASE-8819] - Port HBASE-5428 to Thrift 2
>     [HBASE-8826] - Ensure HBASE-8695 is covered in Thrift 2
>     [HBASE-8832] - Ensure HBASE-4658 is supported by Thrift 2
>     [HBASE-8876] - Addendum to HBASE-8774 Add BatchSize and Filter to
> Thrift2 - Add BatchSize Test
>     [HBASE-8938] - Thrift2 does not close scanner instance
>     [HBASE-8432] - a
>  table with unbalanced regions will balance indefinitely with the
> 'org.apache.hadoop.hbase.master.DefaultLoadBalancer'
>     [HBASE-8678] - Wrongly delete cells in some case which can not be
> deleted
>     [HBASE-8695] - The HBase thrift service ignores XML configuration
>     [HBASE-8776] - tweak retry settings some more (on trunk and 0.94)
>     [HBASE-8782] - Thrift2 can not parse values when using framed transport
>     [HBASE-8794] - DependentColumnFilter.toString() throws
> NullPointerException
>     [HBASE-8811] - REST service ignores misspelled "check=" parameter,
> causing unexpected mutations
>     [HBASE-8814] - Possible NPE in split if a region has empty store files.
>     [HBASE-8858] - Backport hbase-6979 "recovered.edits file should not
> break distributed log splitting"
>     [HBASE-8910] -
>  HMaster.abortNow shouldn't try to become a master again if it was stopped
>     [HBASE-8967] - Duplicate call to snapshotManager.stop() in
> HRegionServer
>     [HBASE-8968] - Avoid call to zk in HRegionServer#getMaster() to log
> the master address which already read from the zk.
>     [HBASE-8971] - Bump 0.94 version to 0.94.10-SNAPSHOT
>     [HBASE-8988] - Reuse the thread pool in the shell to not run out of
> threads
>     [HBASE-8599] - HLogs in ZK are not cleaned up when replication lag is
> minimal
>     [HBASE-8767] - Backport hbase-8001 and hbase-8012, avoid lazy seek
>     [HBASE-8806] - Row locks are acquired repeatedly in