Could not get hold of varnishd

varnish / docker-varnish

Official docker image

https://hub.docker.com/_/varnish

82 stars 34 forks source link

Could not get hold of varnishd #29

Closed octfx closed 3 years ago

octfx commented 3 years ago

Using varnish:fresh or varnish:stable (6.0.7+) no varnish* command inside the docker container works.
Every command outputs Could not get hold of varnishd, is it running?

Using versions 6.0.6 or 6.5 all commands successfully connect to the daemon.

Example compose file used:

version: '3.7'

services:
  varnish:
    image: varnish:fresh
    restart: unless-stopped
    expose:
      - 80

gquintard commented 3 years ago

ah, I get it, you are using varnish* commands after exec'ing in the container right?

We changed things recently so that people could run sidecars more easily: https://github.com/varnish/docker-varnish/commit/3b38b0e1c38c1f6324ac8afa156ff73cfbbffcf4

services:
  varnish:
    image: varnish
    volumes:
      - cache-logs:/var/lib/varnish
  varnish-log:
    image: varnish
    volumes:
      - cache-logs:/var/lib/varnish
    command: [varnishncsa]

volumes:
  cache-logs:
    driver: local
    driver_opts:
      type: tmpfs
      device: tmpfs

but that means by default, varnishd will start with -n /var/lib/varnish. So, in the current situation, you have two options:

run the varnish* commands with -n /var/lib/varnish too so they find the workdir
run the container with the full command line starting with /usr/sbin/varnishd -F -f ... so the entrypoint doesn't inject the argument

I'll check if adding aliases/trampoline scripts could be a better approach to satisfy everyone

octfx commented 3 years ago

This explains and fixes it, thank you!

gquintard commented 3 years ago

latest update to the images includes "trampoline" scripts that are going to inject the -n /var/lib/varnish as first argument before calling the real binary, so now your use case should also work transparently

I'm closing this, but let us know if you you have further issue or question about this

Jancis commented 3 years ago

I stumbled into the same issue with the recent (6.0.x) image. -n does not work for me, but using the full path to binary, does. And I noticed there are two locations for varnishlog, one is symlink to varnishd, another just a binary.

# which varnishlog
/usr/local/bin/varnishlog

# /usr/local/bin/varnishlog -V
/usr/bin/varnishlog (varnish-6.0.7 revision 525d371e3ea0e0c38edd7baf0f80dc226560f26e)
Copyright (c) 2006 Verdens Gang AS
Copyright (c) 2006-2020 Varnish Software AS

# /usr/bin/varnishlog -V
/usr/bin/varnishlog (varnish-6.0.7 revision 525d371e3ea0e0c38edd7baf0f80dc226560f26e)
Copyright (c) 2006 Verdens Gang AS
Copyright (c) 2006-2020 Varnish Software AS

# /usr/local/bin/varnishlog
.....
VSM: Could not get hold of varnishd, is it running?

# /usr/bin/varnishlog
(..)
-   VCL_use        boot
-   Timestamp      Start: 1623244019.731650 0.000000 0.000000
-   BereqMethod    GET
-   BereqURL       /
-   BereqProtocol  HTTP/1.1
(..)

Update: Figured out the issue. I've overridden entrypoint and have not added -n /var/lib/varnish parameter to varnishd. Leaving my comment here, in case anyone else has this issue.