[
https://issues.apache.org/jira/browse/HAWQ-358?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15111380#comment-15111380
]
Caleb Welton commented on HAWQ-358:
-----------------------------------
The gpfdist issues look more significant and are likely due to changes in the
HAWQ architecture in combination with external tables for HAWQ 2.0.
These tests create external table such as:
{noformat}
CREATE EXTERNAL WEB TABLE gpfdist_status (x text)
execute E'( python $GPHOME/bin/lib/gppinggpfdist.py localhost:7070 2>&1 ||
echo) '
on SEGMENT 0
FORMAT 'text' (delimiter '|');
CREATE EXTERNAL WEB TABLE gpfdist_start (x text)
execute E'((/data/hawq-devel/bin/gpfdist -p 7070 -d
/data/hawq/src/test/regress/data </dev/null >/dev/null 2>&1 &); sleep 2; echo
"starting...") '
on SEGMENT 0
FORMAT 'text' (delimiter '|');
CREATE EXTERNAL WEB TABLE gpfdist_stop (x text)
execute E'(/bin/pkill gpfdist || killall gpfdist) > /dev/null 2>&1; echo
"stopping..."'
on SEGMENT 0
FORMAT 'text' (delimiter '|');
{noformat}
With the expectation that the "on SEGMENT 0" specification implies that these
will be run ON SEGMENT 0, aka on the master node. However in the current
implementation these seem to be run on the segments instead, as shown in this
plan:
{noformat}
explain select * from gpfdist_stop;
QUERY PLAN
---------------------------------------------------------------------------------------
Gather Motion 1:1 (slice1; segments: 1) (cost=0.00..11000.00 rows=1000000
width=32)
-> External Scan on gpfdist_stop (cost=0.00..11000.00 rows=1000000
width=32)
(2 rows)
{noformat}
Which means that the test framework is not successfully starting/stopping
gpfdist the way the test framkework intended.
There is also another issue where the framework inserts 'localhost' instead of
the name of the master host into the external table definitions.
Manually starting gpfdist and using the right hostname results in the gpfdist
external tables working correctly.
So in summary:
* We have a bug in the handling of the "on segment X" syntax for external tables
* We have a test framework issue setting the gpfdist hostname correctly.
> Installcheck good failures in hawq-dev environment
> --------------------------------------------------
>
> Key: HAWQ-358
> URL: https://issues.apache.org/jira/browse/HAWQ-358
> Project: Apache HAWQ
> Issue Type: Bug
> Components: Tests
> Reporter: Caleb Welton
> Assignee: Jiali Yao
>
> Build and test within a hawq dev environment setup via the instructions
> outlined in the hawq-devel docker enviroment:
> https://hub.docker.com/r/mayjojo/hawq-devel/
> Results in the following errors
> {noformat}
> ...
> test errortbl ... FAILED (6.83 sec)
> ...
> test subplan ... FAILED (8.15 sec)
> ...
> test create_table_distribution ... FAILED (3.47 sec)
> test copy ... FAILED (34.76 sec)
> ...
> test set_functions ... FAILED (4.90 sec)
> ...
> test exttab1 ... FAILED (17.66 sec)
> ...
> {noformat}
> Summary of issues:
> * *errortbl* - every connection to gpfdist results in "connection with
> gpfdist failed for gpfdist://localhost:7070/nation.tbl"
> * *subplan* - trying to create plpython resulted in "could not access file
> "$libdir/plpython": No such file or directory", lack of plpython causes many
> other statements to fail
> * *create_table_distribution* - test likely needs some refactoring to reflect
> calculating correct bucketnum based on current system configuration
> * *copy* - seems to be failing because rows aren't coming out in the expected
> order, test needs fixing to be able to handle this
> * *set_functions* - same plpythonu issue described above
> * *exttab1* - same issue reading from gpfdist described above
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
