[
https://issues.apache.org/jira/browse/IMPALA-8552?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16841793#comment-16841793
]
David Knupp edited comment on IMPALA-8552 at 5/16/19 11:00 PM:
---------------------------------------------------------------
[~tarmstrong] – I was confused by this too. I think one of the problems is that
our build scripts and our test code are too intertwined. There are too many
points of entry, and it's nearly impossible to track all of it – especially
since we tend to rely too heavily on arbitrary global env variables. In this
case, the code you're referencing was introduced relatively recently in
[https://github.com/apache/impala/commit/691f9d9].
In this case, is_remote_cluster() seems to rely on IMPALA_REMOTE_URL, which was
added as part of the same commit.
{noformat}
$ git grep IMPALA_REMOTE_URL
tests/common/environ.py:IMPALA_REMOTE_URL = os.environ.get("IMPALA_REMOTE_URL",
"")
{noformat}
Elsewhere, an env variable called REMOTE_LOAD is used to determine whether data
load scripts are loading script locally or not.
Finally, in conftest.py, there's also a flag for {{--testing_remote_cluster}}
(and a SkipIf definition based upon it). There's nothing that coordinates any
of these inputs, or guarantees that they need to agree. It's a little maddening.
was (Author: dknupp):
I was confused by this too. I think one of the problems is that our build
scripts and our test code are too intertwined. There are too many points of
entry, and it's nearly impossible to track all of it -- especially since we
tend to rely too heavily on arbitrary global env variables. In this case, the
code you're referencing was introduced relatively recently in
https://github.com/apache/impala/commit/691f9d9.
In this case, is_remote_cluster() seems to rely on IMPALA_REMOTE_URL, which was
added as part of the same commit.
{noformat}
$ git grep IMPALA_REMOTE_URL
tests/common/environ.py:IMPALA_REMOTE_URL = os.environ.get("IMPALA_REMOTE_URL",
"")
{noformat}
Elsewhere, an env variable called REMOTE_LOAD is used to determine whether data
load scripts are loading script locally or not.
Finally, in conftest.py, there's also a flag for {{--testing_remote_cluster}}
(and a SkipIf definition based upon it). There's nothing that coordinates any
of these inputs, or guarantees that they need to agree. It's a little maddening.
> impala-shell tests break on deployed clusters if IMPALA_LOCAL_BUILD_VERSION
> is None
> -----------------------------------------------------------------------------------
>
> Key: IMPALA-8552
> URL: https://issues.apache.org/jira/browse/IMPALA-8552
> Project: IMPALA
> Issue Type: Bug
> Components: Infrastructure
> Affects Versions: Impala 3.3.0
> Reporter: David Knupp
> Assignee: Tim Armstrong
> Priority: Critical
>
> This is a regression introduced by the commit:
> https://github.com/apache/impala/commit/b55d905
> *Stacktrace*
> {noformat}
> shell/test_shell_commandline.py:33: in <module>
> from util import (get_impalad_host_port, assert_var_substitution,
> run_impala_shell_cmd,
> shell/util.py:42: in <module>
> IMPALA_HOME, "shell/build", "impala-shell-" + IMPALA_LOCAL_BUILD_VERSION,
> E TypeError: cannot concatenate 'str' and 'NoneType' objects
> {noformat}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
