1

We are using titan 0.5.0 over hbase 0.98.6 (cloudera 5.2). After some time of use we find out one column familiy in a specific region is bigger than the HBase Maximum File Size (hbase.hregion.max.filesize) and yet whenever we split the region it stay the same size.

Anyone know this problem and how we can fix it?

Nimrod
  • 370
  • 2
  • 11

2 Answers2

0

We realized the problem was with one of the indexes. There was an index over a low-cardinality property. That made a single index row in the hbase to be very large. We couldnt split it because it was just a single row.

Nimrod
  • 370
  • 2
  • 11
0

hbase.hregion.max.filesize is not work from hbase 0.94+ ,because default split policy is changed. check this link

you should change split_policy in shell, like this:{TABLE_ATTRIBUTES => {MAX_FILESIZE => '12884901888', CONFIGURATION => {'hbase.hregion.scan.loadColumnFamiliesOnDemand' => 'true', 'SPLIT_POLICY' => 'org.apache.hadoop.hbase.regionserver.ConstantSizeRegionSplitPolicy'}}

Community
  • 1
  • 1
dape
  • 565
  • 1
  • 8
  • 9