Jan Høydahl created SOLR-18647:
----------------------------------
Summary: bin/solr.cmd silently ignores "stop --all" whenever
SOLR_PORT_LISTEN is set
Key: SOLR-18647
URL: https://issues.apache.org/jira/browse/SOLR-18647
Project: Solr
Issue Type: Bug
Components: cli
Reporter: Jan Høydahl
*Found by CATS*, the cross-platform CLI test framework proposed in SOLR-18642 —
see [PR #5323|https://github.com/apache/solr/pull/5323]. Two tests there expose
this, and will be marked {{@AwaitsFix}} against this issue so the suite stays
green. A fix can land before CATS is merged or after; if it lands first, please
also drop the {{@AwaitsFix}} annotation from those two tests on that PR.
h3. The bug
{{bin/solr.cmd}} nests the whole {{--all}} implementation inside a test for the
port being unset:
{code}
826: IF "%SOLR_PORT_LISTEN%"=="" (
827: IF "%STOP_ALL%"=="1" (
... walks %SOLR_PID_DIR%, stops every node it finds ...
860: if "!found_it!"=="0" echo No Solr nodes found to stop.
)
865: ) ELSE (
... stops exactly %SOLR_PORT_LISTEN% ...
887: if "!found_it!"=="0" echo No Solr found running on port
%SOLR_PORT_LISTEN%
{code}
So when a port is in effect, {{--all}} is parsed, accepted, and then ignored:
control goes to the single-port branch and {{STOP_ALL}} is never read again.
The user is not told.
{{bin/solr}} has no such condition — its {{--all}} branch at line 951 ({{if
$stop_all; then}}) runs whenever the command is {{stop}}, so the two scripts
disagree about precedence.
h3. Why this is the ordinary case, not an exotic one
{{SOLR_PORT_LISTEN}} is a documented setting in {{solr.in.cmd}}, and it also
comes from the environment and from {{-p}}. Anyone who pins the port that way
loses {{solr stop \-\-all}} entirely, silently.
h3. Evidence
>From the first full run of the CATS suite under cmd.exe:
{{StartSolrTest.solr11740CheckSolrStopConnection}} starts nodes on 22158 and
22159, then {{solr stop \-\-all}} exits 0 having printed only {{Stopping Solr
process 7116 running on port 22158}} — node 22159 is still running afterwards.
{{\-\-all}} stopped one node of two.
{{NoPidFileTest.stopAllReportsWhenItStoppedNothing}} runs {{solr stop \-\-all}}
with nothing running and gets {{No Solr found running on port 22150}}, the
single-port message at line 887. The global {{No Solr nodes found to stop.}} at
line 860 is unreachable, which is why the assertion fails even though
{{solr.cmd}} contains that exact string.
h3. Secondary, not worth its own issue
The two scripts word a graceful stop differently: {{bin/solr}} prints {{Sending
stop command}}, {{solr.cmd}} prints {{Stopping Solr process N running on port
P}}. A parity nit, worth folding in if a fix touches those lines.
🤖 AI Assisted
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
