[CASSANDRA-6239] pidfile is never written, "/etc/init.d/cassandra stop" fails

• Type: Bug
• Status: Resolved
• Priority: Normal
• Resolution: Duplicate
• Component/s: Packaging
• Severity: Normal
• Assignee: Unassigned Assign to me
• Reporter: Faidon Liambotis
• Created: 24/Oct/13 20:40
• Updated: 16/Apr/19 09:32
• Resolved: 24/Oct/13 22:07

Description

The init script tries, via start-stop-daemon, to write a pidfile to /var/run/cassandra/cassandra.pid. /var/run/cassandra doesn't exist (righftully so, /var/run can be a tmpfs), so the init script has this stanza above the start-stop-daemon invocation:

    [ -e `dirname "PIDFILE"` ] || \
        install -d -ocassandra -gcassandra -m750 `dirname $PIDFILE`

The first line is missing the dollar sign before the PIDFILE variable (i.e. it should be 'dirname "$PIDFILE"'). This has the effect that "dirname PIDFILE" is called, with the PIDFILE as a literal, which always returns "." as the output, which always exists, so the "install" call never gets executed, the directory never gets created and start-stop-daemon is unable to write the pidfile.

The pidfile is never written and "/etc/init.d/cassandra stop" never works.

Adding a $ before PIDFILE fixes the issue. This has been tested.

Issue Links

• duplicates: CASSANDRA-6101 Debian init script broken