KUDU-602
add addCallbacksDeferring to async hbaseKUDU-639
Leader doesn't overwrite demoted follower's log properly
Add metrics for LRU cache and IO
Crash between creating log metadata and log container prevents restartKUDU-562
When a peer is partitioned, it should ask a majority if they still think the leader is alive
On Log appends, allow to specify if an fsync is required per message
full stack "with disk" test doesn't wait for flushes
Don't trigger elections when waiting for our own Log
Better timeout handling for Kudu clients, especially for Master requests
Make LogCache more flexible with regard to what it keeps in memory
Implement multi-disk support
Use SleepFor(MonoDelta) rather than sleep or usleep
Master service doesn't set the required code field of MasterErrorPB
Ensure that log block manager punched holes actually reclaim space
Block storage operability improvements
Enable failure detection/leader election by default
"Cannot allocate memory" error in long running testKUDU-106
Rename WAL files from log- to wal-
Implement data block GC
Operability improvements based on cluster testing
Prototype log container allocation strategy
Integrate new block storage API with Kudu
Reimplement current Kudu on-disk storage approach as allocation strategy
Design API for block storage layer
Avoid blocking in KuduClient::Data::SetMasterServerProxyAsync
Tablet startup should not eagerly open every CFile
major delta compaction should generate UNDOs
Integrate maintenance ops in master
c++: CreateTable polls IsCreateTableDone() in a tight loop
CPU performance bottlenecks in block cache
Failed compilation of row projector from base schema Schema
Figure out how to distribute client library
Deadlock between Consensus shutdown and Peer response
Log updates last_entry_op_id_ before it is durable
On boot make sure to discard the prefix of segments without operations with ids
c++ client: Support for configurable timeouts on scanners
Test when errors occur on the Apply() phase of transactions
fsync parent directory when flushing data files
Replicate and Prepare can become mis-ordered during replica->leader transition
Fix uses of fsync vs. fdatasync
long LLT run fails on snapshot verification
CHECK failure in DiskRowSetWriter::AppendBlock() when you get a delta flush during the middle of a compaction
Crash if multiple updates to a flushed row come in the same batch