Joe McDonnell created IMPALA-12195:
--------------------------------------
Summary: bin/kill-ranger-server.sh's kill of Ranger can fail with
"Operation not permitted"
Key: IMPALA-12195
URL: https://issues.apache.org/jira/browse/IMPALA-12195
Project: IMPALA
Issue Type: Bug
Components: Infrastructure
Affects Versions: Impala 4.3.0
Reporter: Joe McDonnell
Some Jenkins jobs running Impala tests have hit an error when running
testdata/bin/kill-ranger-server.sh:
{noformat}
19:25:18 Getting pid from /tmp/rangeradmin.pid ..
19:25:20 Found Apache Ranger Admin Service with pid 20557, Stopping it...
19:25:20
{dirname}/Impala-Toolchain/cdp_components-40643771/ranger-2.3.0.7.2.16.0-287-admin/ews/ranger-admin-services.sh:
line 129: {dirname}/logs/cluster/ranger/catalina.out: No such file or directory
19:25:20 Shutdown in progress. Will check after 2 secs again..
19:25:22 Shutdown in progress. Will check after 2 secs again..
19:25:24 Shutdown in progress. Will check after 2 secs again..
19:25:26 Shutdown in progress. Will check after 2 secs again..
19:25:28 Shutdown in progress. Will check after 2 secs again..
19:25:30 Shutdown in progress. Will check after 2 secs again..
19:25:32 Shutdown in progress. Will check after 2 secs again..
19:25:34 Shutdown in progress. Will check after 2 secs again..
19:25:36 Shutdown in progress. Will check after 2 secs again..
19:25:38 Shutdown in progress. Will check after 2 secs again..
19:25:40 Shutdown in progress. Will check after 2 secs again..
19:25:42 Shutdown in progress. Will check after 2 secs again..
19:25:44 Shutdown in progress. Will check after 2 secs again..
19:25:46 Shutdown in progress. Will check after 2 secs again..
19:25:48 Shutdown in progress. Will check after 2 secs again..
19:25:50 Initial kill failed, getting serious now...
19:25:50
{dirname}/Impala-Toolchain/cdp_components-40643771/ranger-2.3.0.7.2.16.0-287-admin/ews/ranger-admin-services.sh:
line 143: kill: (20557) - Operation not permitted
19:25:50 Wow, even kill -9 failed, giving up! Sorry..{noformat}
The job in question has never started Ranger. This is calling
kill-ranger-server.sh via testdata/bin/kill-all.sh from testdata/bin/run-all.sh
as part of starting the minicluster.
One possibility is that the /tmp/rangeradmin.pid is leftover from an old run
and the pid has been reused. The ranger-admin-services.sh script doesn't seem
to remove the pid file after stopping Ranger.
If this is the cause, then we can customize the location of the Ranger pid file
and make sure it gets deleted properly.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
