Outline:

We are trying to connect up varnish-4.1.11 to magento 1 in kubernetes using the nexcess turpentine addon, but the same error is returned each time:

Error determining Varnish version: Varnish admin socket timeout
Failed to load configurator

Application stack:

We have a kubernetes cluster running a magento 1 stack with the following containers:

• php-fpm:7.2/nginx:latest
• mysql:5.7
• redis:latest
• nfs-provisioner:latest
• nginx:latest (acts as a proxy for varnish to point to)
• varnish:4.1.11

kubernetes info:

• Networking: cilium:v16.3

Server Version: version.Info{Major:"1", Minor:"15", GitVersion:"v1.15.5", GitCommit:"20c265fef0741dd71a66480e35bd69f18351daea", GitTreeState:"clean", BuildDate:"2019-10-15T19:07:57Z", GoVersion:"go1.12.10", Compiler:"gc", Platform:"linux/amd64"}

Varnish config:

NFILES=131072
MEMLOCK=82000
NPROCS="unlimited"
RELOAD_VCL=1
VARNISH_VCL_CONF=/var/www/html/site/var/default.vcl
VARNISH_LISTEN_PORT=6081
VARNISH_ADMIN_LISTEN_PORT=6082
VARNISH_SECRET_FILE=/etc/varnish/secret
VARNISH_MIN_THREADS=5
VARNISH_MAX_THREADS=50
VARNISH_THREAD_TIMEOUT=120
VARNISH_STORAGE="malloc,512M"
VARNISH_TTL=120

DAEMON_OPTS="-F -a ${VARNISH_LISTEN_ADDRESS}:${VARNISH_LISTEN_PORT} \
             -f ${VARNISH_VCL_CONF} \
             -T ${VARNISH_ADMIN_LISTEN_ADDRESS}:${VARNISH_ADMIN_LISTEN_PORT} \
             -t ${VARNISH_TTL} \
             -S ${VARNISH_SECRET_FILE} \
             -s ${VARNISH_STORAGE}" \
             -p esi_syntax=0x2 \
             -p cli_buffer=16384

What we've tried so far:

• Downgrading to varnish-3.0.7
• Pointing magento to varnish's IP directly
• Running a generic varnish connection script in PHP

Notes:

• Pinging the varnish pod from the nginx/fpm pod works fine
• Curling to the varnish ports from the nginx/fpm pod also works fine
• The generic connection script noted above works successfully when run from inside the varnish container itself, which very likely indicates a networking issue.
• Running the stack locally in docker-compose works fine, which also indicates a networking issue.

I appreciate that this is a very very niche issue, but hopefully someone else has some insight into what could be going wrong.