[PHOENIX-4630] Reverse scan with row_timestamp does not working

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 4.13.0, 4.13.1, 4.13.2, 4.13.2-cdh
• Fix Version/s: 4.14.0
• Assignee: chenglei
• Reporter: JeongMin Ju
• Created: 28/Feb/18 01:22
• Updated: 02/Mar/18 03:19
• Resolved: 02/Mar/18 03:19

Description

In version 4.13, if the query plan is a reverse scan, an error or incorrect data is returned.

This is a problem that occurs when the query plan is a reverse range scan in the case of an "order by desc" query for the row key.

The table schem is as follows.

create table if not exists app_log (
  app_tag varchar not null,
  timestamp date not null,
  uuid varchar not null,
  log varchar
  constraint pk primary key(app_tag, timestamp row_timestamp, uuid)
)
data_block_encoding='FAST_DIFF',
compression='LZ4',
update_cache_frequency=600000,
column_encoded_bytes = 1,
ttl=2592000,
salt_buckets=50
;

The current data is as follows.

upsert into app_log values ('test', now(), 'test', 'test');
...

select * from app_log order by timestamp;
+-----------+--------------------------+-------+-------+
|  APP_TAG  |        TIMESTAMP         | UUID  |  LOG  |
+-----------+--------------------------+-------+-------+
| test      | 2018-02-28 01:02:16.985  | test  | test  |
| test      | 2018-02-28 01:02:19.472  | test  | test  |
| test      | 2018-02-28 01:02:21.568  | test  | test  |
| test      | 2018-02-28 01:02:23.332  | test  | test  |
| test      | 2018-02-28 01:02:25.200  | test  | test  |
| test      | 2018-02-28 01:02:27.055  | test  | test  |
| test      | 2018-02-28 01:02:29.008  | test  | test  |
| test      | 2018-02-28 01:02:30.911  | test  | test  |
| test      | 2018-02-28 01:02:32.775  | test  | test  |
| test      | 2018-02-28 01:02:34.663  | test  | test  |
+-----------+--------------------------+-------+-------+

You can see errors if you run a simple query after adding some data.

Depending on the data, an error may occur and incorrect data may be output.

select * from app_log where app_tag = 'test' and timestamp between to_date('2018-02-28 01:02:16') and to_date('2018-02-28 01:02:34') order by timestamp desc;

Error: org.apache.phoenix.exception.PhoenixIOException: org.apache.hadoop.hbase.DoNotRetryIOException: APP_LOG,\x0D\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00,1519778082466.6dd30a7d7a26a38c5c06d63008bbff3d.: seekToPreviousRow must not be called on a non-reversed scanner
at org.apache.phoenix.util.ServerUtil.createIOException(ServerUtil.java:96)
at org.apache.phoenix.util.ServerUtil.throwIOException(ServerUtil.java:62)
at org.apache.phoenix.iterate.RegionScannerFactory$1.nextRaw(RegionScannerFactory.java:212)
at org.apache.phoenix.coprocessor.DelegateRegionScanner.nextRaw(DelegateRegionScanner.java:82)
at org.apache.phoenix.coprocessor.DelegateRegionScanner.nextRaw(DelegateRegionScanner.java:82)
at org.apache.phoenix.coprocessor.BaseScannerRegionObserver$RegionScannerHolder.nextRaw(BaseScannerRegionObserver.java:293)
at org.apache.hadoop.hbase.regionserver.RSRpcServices.scan(RSRpcServices.java:2561)
at org.apache.hadoop.hbase.protobuf.generated.ClientProtos$ClientService$2.callBlockingMethod(ClientProtos.java:33648)
at org.apache.hadoop.hbase.ipc.RpcServer.call(RpcServer.java:2183)
at org.apache.hadoop.hbase.ipc.CallRunner.run(CallRunner.java:112)
at org.apache.hadoop.hbase.ipc.RpcExecutor$Handler.run(RpcExecutor.java:185)
at org.apache.hadoop.hbase.ipc.RpcExecutor$Handler.run(RpcExecutor.java:165)
Caused by: org.apache.commons.lang.NotImplementedException: seekToPreviousRow must not be called on a non-reversed scanner
at org.apache.hadoop.hbase.regionserver.NonReversedNonLazyKeyValueScanner.seekToPreviousRow(NonReversedNonLazyKeyValueScanner.java:44)
at org.apache.hadoop.hbase.regionserver.ReversedKeyValueHeap.seekToPreviousRow(ReversedKeyValueHeap.java:89)
at org.apache.hadoop.hbase.regionserver.ReversedRegionScannerImpl.nextRow(ReversedRegionScannerImpl.java:71)
at org.apache.hadoop.hbase.regionserver.HRegion$RegionScannerImpl.nextInternal(HRegion.java:5938)
at org.apache.hadoop.hbase.regionserver.HRegion$RegionScannerImpl.nextRaw(HRegion.java:5673)
at org.apache.hadoop.hbase.regionserver.HRegion$RegionScannerImpl.nextRaw(HRegion.java:5659)
at org.apache.phoenix.iterate.RegionScannerFactory$1.nextRaw(RegionScannerFactory.java:175)
... 9 more (state=08000,code=101)

 Query plan is as follow.

explain select * from app_log where app_tag = 'test' and timestamp between to_date('2018-02-28 01:02:20') and to_date('2018-02-28 01:02:30') order by timestamp desc;
+------------------------------------------------------------------------------------------------------------------------------------------------------------+-----------------+----------------+--------------+
|                                                                            PLAN                                                                            | EST_BYTES_READ  | EST_ROWS_READ  | EST_INFO_TS  |
+------------------------------------------------------------------------------------------------------------------------------------------------------------+-----------------+----------------+--------------+
| CLIENT 14-CHUNK PARALLEL 14-WAY REVERSE RANGE SCAN OVER APP_LOG [0,'test','2018-02-28 01:02:20.000'] - [49,'test','2018-02-28 01:02:30.000']  | null            | null           | null         |
|     ROW TIMESTAMP FILTER [1519779740000, 1519779750001)                                                                                                    | null            | null           | null         |
| CLIENT MERGE SORT                                                                                                                                          | null            | null           | null         |
+------------------------------------------------------------------------------------------------------------------------------------------------------------+-----------------+----------------+--------------+

Notice that the query plan is REVERSE RANGE SCAN.

Refer to my comment on PHOENIX-4622 for causes and solutions.

Issue Links

• is caused by: PHOENIX-4277 Treat delete markers consistently with puts for point-in-time scans