custom pidfile causing restart to fail

[mgla]

Steps to reproduce:

Fire up a new Ubuntu 14.04 instance (I used an AWS Ubuntu image):

add-apt-repository "deb https://repo.varnish-cache.org/ubuntu/ trusty varnish-4.1"
apt-get update
apt-get install varnish --force-yes -y
sed -i 's#-s malloc,256m"#-P /run/varnishd.pid \\\n             -s malloc,256m"#' /etc/default/varnish
service varnish restart

Expected behavior

Clean restart

Actual behavior

 * Stopping HTTP accelerator varnishd
No /usr/sbin/varnishd found running; none killed.
   ...fail!
 * Starting HTTP accelerator varnishd
   ...fail!
Error: Cannot open socket: :6081: Address already in use

Additional details:

The sed just adds the -P parameter to /etc/default/varnish:

DAEMON_OPTS="-a :6081 \
             -T localhost:6082 \
             -f /etc/varnish/default.vcl \
             -S /etc/varnish/secret \
             -P /run/varnishd.pid \
             -s malloc,256m"

Took me a while to track it down. I recently opened this report: #71 but was wrong in my assessment.

[mgla]

Pinging @denisbr I managed to reproduce it, but it was not what I expected.

[denisbr]

Does the varnish use have sufficient privileges on the /run directory?

[denisbr]

Also, could you dump the complete resulting /etc/default/varnish?

[mgla]

Complete /etc/default/varnish:

# Configuration file for Varnish Cache.
#
# /etc/init.d/varnish expects the variables $DAEMON_OPTS, $NFILES and $MEMLOCK
# to be set from this shell script fragment.
#
# Note: If systemd is installed, this file is obsolete and ignored.  You will
# need to copy /lib/systemd/system/varnish.service to /etc/systemd/system/ and
# edit that file.

# Should we start varnishd at boot?  Set to "no" to disable.
START=yes

# Set WARMUP_TIME to force a delay in reload-vcl between vcl.load and vcl.use
# This is useful when backend probe definitions need some time before declaring
# configured backends healthy, to avoid routing traffic to a non-healthy backend.
#WARMUP_TIME=0

# Maximum number of open files (for ulimit -n)
NFILES=131072

# Maximum locked memory size (for ulimit -l)
# Used for locking the shared memory log in memory.  If you increase log size,
# you need to increase this number as well
MEMLOCK=82000

DAEMON_OPTS="-a :6081 \
             -T localhost:6082 \
             -f /etc/varnish/default.vcl \
             -S /etc/varnish/secret \
             -P /run/varnishd.pid \
             -s malloc,256m"

[mgla]

/run/varnishd.pid is actually the default pid file. @denisbr This might not even be an issue with the packaging. Should I reopen an issue at the varnish repository? https://github.com/varnishcache/varnish-cache

[denisbr]

This is definitely not a varnish-cache issue, and while it is caused by packaging and the initscripts, it's not really a bug per-se.

The issue is that the initscript has it's own pidfile handling:
https://github.com/varnishcache/pkg-varnish-cache/blob/master/debian/varnish.init#L23-L52

If you want to use an alternative pidfile path / filename, the correct way to do this in debian is to add a PIDFILE option to the /etc/default/varnish file.

PIDFILE="/run/mycustompidfile.pid

This will override the predefined PIDFILE in the initscript, and make all interaction with the script work as expected.

[mgla]

@denisbr Actually, the init script ignores externally set PIDFILE and hardcodes it to /run/varnishd.pid, see /etc/init.d/varnish

Edit 1: I can still reproduce the behavior when I use PIDFILE in /etc/default/varnish, and will investigate. Please do not close the issue just yet.

Edit 2: You are right, nevermind.

[denisbr]

@mgla I did the work to test all scenarios that I mentioned in local environments before closing.

[mgla]

@denisbr You were absolutely right. I believieve this can be closed.

[denisbr]

Thanks for confirming, leaving the issue closed :)