[
https://issues.apache.org/jira/browse/PHOENIX-3920?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Samarth Jain updated PHOENIX-3920:
----------------------------------
Attachment: PHOENIX-3920_4.x-HBase-0.98.patch
Patch that takes care of creating a guide post for the last chunk that was read
by our statistics collector. We weren't always creating a guide post for this
last remaining chunk because the chunk size was less than the guide post size.
To verify my fix, I took the data set that [~mujtabachohan] used for
PHOENIX-3836. Earlier the explain used to report the number of rows scanned as
994728 although the entire table of 1 million rows was being scanned.
{code}
+--------------------------------------------------------------------------------+-----------------+----------------+
| PLAN
| EST_BYTES_READ | EST_ROWS_READ |
+--------------------------------------------------------------------------------+-----------------+----------------+
| CLIENT 161-CHUNK 994728 ROWS 1667243454 BYTES PARALLEL 1-WAY FULL SCAN OVER F
| 1667243454 | 994728 |
| SERVER FILTER BY FIRST KEY ONLY
| 1667243454 | 994728 |
| SERVER AGGREGATE INTO SINGLE ROW
| 1667243454 | 994728 |
+--------------------------------------------------------------------------------+-----------------+----------------+
{code}
With this patch, the number of rows scanned are now collected and reported
correctly.
{code}
+---------------------------------------------------------------------------------+-----------------+----------------+
| PLAN
| EST_BYTES_READ | EST_ROWS_READ |
+---------------------------------------------------------------------------------+-----------------+----------------+
| CLIENT 163-CHUNK 1000000 ROWS 1676080378 BYTES PARALLEL 1-WAY FULL SCAN OVER
F | 1676080378 | 1000000 |
| SERVER FILTER BY FIRST KEY ONLY
| 1676080378 | 1000000 |
| SERVER AGGREGATE INTO SINGLE ROW
| 1676080378 | 1000000 |
+---------------------------------------------------------------------------------+-----------------+----------------+
{code}
[~jamestaylor], please review.
> Stats collection doesn't always create a guide post for last remaining chunk
> ----------------------------------------------------------------------------
>
> Key: PHOENIX-3920
> URL: https://issues.apache.org/jira/browse/PHOENIX-3920
> Project: Phoenix
> Issue Type: Bug
> Reporter: Samarth Jain
> Assignee: Samarth Jain
> Fix For: 4.11.0
>
> Attachments: PHOENIX-3920_4.x-HBase-0.98.patch
>
>
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
