metabase 🚀 - Allow Metabase to run at a non-root app context such as /mb/

No, there is currently no way to configure Metabase to run at a non-root app context

agilliland on 5 Mar 2016

A possible simple fix for this would be to use the value (or at least the path portion?) of the site-url setting in a <base href="..."> tag of the <head> of index.html, and use relative instead of absolute URLs for everything.

https://stackoverflow.com/questions/1889076/is-it-recommended-to-use-the-base-html-tag

tlrobinson on 5 Mar 2016

@agilliland Is this up for a pull request? Or, not a priority right now. I may be able to contribute.

Jacob-Kroeze on 10 Mar 2016

sure, if you want to take a stab you are welcome to go for it. i'd encourage doing a quick write up of how you plan to proceed and sharing it to get feedback before doing too much coding.

agilliland on 10 Mar 2016

@Jacob-Kroeze I took a stab at this in #2393. Still very experimental (some images and probably other things are broken)

tlrobinson on 21 Apr 2016

@tlrobinson Hi, I find this feature quite useful. Can I somehow help you with your PR? Do you know why is your branch failing to build?

TheMP on 12 May 2016

+1 for this feature request. I would like to configure a friendly URL like http://mycompanyserver/analytics/ to point to metabase for our enterprise.

lxchavez on 25 May 2016

We have same issue as lxchavez in my company as well.
Small but useful feature

moumny on 8 Jun 2016

👍3

So I was trying to set it up the way @tlrobinson has suggested, and after doing some changes and pushing it up, it gives me following error on browser.

Resource interpreted as Stylesheet but transferred with MIME type text/html: "http://104.198.196.198/metabase/app/dist/styles.bundle.css?4a405a2abca714f472df06e69d1399ff"

Code: https://github.com/VikramTiwari/metabase/tree/feature_non_root_app
Live: http://104.198.196.198/metabase/

VikramTiwari on 26 Jul 2016

Can someone help me find out what am I missing here?

VikramTiwari on 26 Jul 2016

I'm not sure exactly what's going on there. What webserver + configuration are you using in front of Metabase?

As I mentioned above, that branch is experimental. At this points it's diverged a lot from masterand will likely need to be started from scratch.

tlrobinson on 26 Jul 2016

So I have started from current master (as of yesterday), and updated code is here. The webserver is nginx and with following configuration:

server {
        listen 80;
        listen [::]:80;
        server_name 104.198.196.198;
        location /metabase/ {
                proxy_pass http://127.0.0.1:8080;
        }
}

Inside the server the code is running in a docker container built using dockerfile present in the project.

VikramTiwari on 26 Jul 2016

FYI if anyone wants to attempt this now you'll probably want to take a look at the discussions here now that we're using react-router: https://github.com/reactjs/react-router/issues/353

tlrobinson on 9 Aug 2016

@tlrobinson do you mean it is possible with a small change to solve this? what would you add/change to allow me to run off /mb/?

I can inject changes using nginx sub_filter directive, that replaces text in files as they are served. So I can fpr instance transform all '="/app/' into '="/mb/app'. It's the react.js router that's giving me trouble...

dncpax on 31 Aug 2016

@dncpax I suspect it would be difficult to use sub_filter to perfectly replace every URL. There are a number of types of URLs we need to make sure get replaced:

links: i.e. <a href="...">s which should now all be using react-router's <Link to="...">, which will make this easy. My last comment was just referring to this case.
JS/CSS/image resources: most of these are handled by webpack now, which will also make this pretty easy (my first attempt had trouble with images referenced in .css files but that should be solvable)
API endpoints: previously I just made all of these URLs relative and injected a <base href="..."> tag with the root path. Not sure if that's the best way or not.
misc: things like window.location = ..., history.pushState(...) etc

tlrobinson on 6 Sep 2016

@tlrobinson well I have it replacing all requests already, so I can see on chrome console all requests coming from /mb/.

Now, if I don't add a then nothing is displayed because an error occurs (Cannot read property className of undefined in angular.min.js:93).
If I do add the base href tag, then I can see all content. Only login does not work ;)
I can login using url:3000 and continue browsing /mb/ just fine.

But as soon as I start browsing cards everything breaks apart... this is where I can see the router going crazy...

dncpax on 7 Sep 2016

works for us, in case it can help other: mind you we using SSL, terminated by nginx

upstream metabase_server {
server 127.0.0.1:3000 fail_timeout=0;
}

server {
listen 80;
server _;
return 301 https:://metabase.xxx.com:$request_uri;
}

server {
listen 443;
server_name metabase.xxx.com;

include /etc/nginx/modules.d/ssl.conf;

#pass through headers from metabase which are considered invalid by Nginx server.
ignore_invalid_headers off; 
location @metabase {
    sendfile off;
    proxy_pass      http://metabase_server;
    proxy_redirect  default;

    proxy_set_header   Host             $host;
    proxy_set_header   X-Real-IP        $remote_addr;
    proxy_set_header   X-Forwarded-For  $proxy_add_x_forwarded_for;
    proxy_max_temp_file_size 0;

    # this is the maximum upload size
    client_max_body_size    10m;
    client_body_buffer_size 128k;

    proxy_connect_timeout   90;
    proxy_send_timeout      90;
    proxy_read_timeout      90;

    proxy_buffer_size           4k;
    proxy_buffers               4 32k;
    proxy_busy_buffers_size     64k;
    proxy_temp_file_write_size  64k;
}

location / {
    try_files $uri @metabase;
}

}

luc-poynt on 16 Feb 2017

@luc-poynt Thanks! Will try it out when time allows. Looks promissing!

dncpax on 16 Feb 2017

sub_filter probably only will work for assets, not quite for api endpoints (if someone succeeds, do let me know).

In the end, messing with the content of a response is just asking for trouble, even for a good proxy like nginx.

deitch on 27 Feb 2017

This would be really useful.

At the moment, this is a deal breaker for us. None of the suggested workarounds are useable in our context (auto-provisioning of a lot of apps behind a proxy on deployment).

maesenka on 23 Mar 2017

@luc-poynt with your setup what is the metabase url our users use?

dncpax on 24 Mar 2017

@dncpax with my init script it runs as the user metabase

see the attache init script we use

init-metabase.zip

luc-poynt on 24 Mar 2017

@luc-poynt what I meant was what does your metabase url look like? What url do your enter on the browser?

dncpax on 29 Mar 2017

https://metabase.domain.com/

luc-poynt on 29 Mar 2017

ok, now I get it. I was trying something like https://server/metabase...

dncpax on 29 Mar 2017

@luc-poynt you have a typo in your init script. line 183:
function relaod_it()

dncpax on 31 Mar 2017

😄1

Im traveling, but I think https://server/metabase is doable, just need couple rewrite, is there any info as what url is breaking ?

here some ES/Kibana ngisn config : we would need something very similar

# Kibana web interface
# This is the default behavior, go to kibana
location / {
    proxy_pass http://localhost:5601/;
    proxy_http_version 1.1;
    proxy_set_header Host $host;
    proxy_set_header X-Real-IP $remote_addr;
    proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
    proxy_pass_request_headers on;
    proxy_connect_timeout 150;
    proxy_send_timeout 100;
    proxy_read_timeout 100;
    proxy_buffers 4 32k;
    client_max_body_size 8m;
    client_body_buffer_size 128k;
}

# Kibana web interface
location ~ ^/kibana/(?<kb>.*) {
    proxy_set_header X-Real-IP $remote_addr;
    proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
    proxy_set_header Host $http_host;
    proxy_set_header X-NginX-Proxy true;
    proxy_set_header Upgrade $http_upgrade;
    proxy_set_header Connection "upgrade";
    proxy_read_timeout 900;
    proxy_connect_timeout 900;
    proxy_pass http://127.0.0.1:5601/$kb;
    proxy_redirect off;
}

# elasticsearch and kopf plugin interface
location ~ ^/es.*$ {
    proxy_pass http://127.0.0.1:9200;
    rewrite ^/es(.*) /$1 break;
}

location ~ ^/kopf/.*$ {
    proxy_pass http://localhost:9200;
    rewrite ^/kopf/(.*) /_plugin/kopf/$1 break;
}

location ~ ^/head/.*$ {
    proxy_pass http://localhost:9200;
    rewrite ^/head/(.*) /_plugin/head/$1 break;
}

luc-poynt on 31 Mar 2017

@luc-poynt I am very interested on making this work. I'll try revive my config and see what doesn't work. But looking at this thread it seems I got to the point where login and cards didn't work...
I don't know proxying rules enough to fully understand your config...

dncpax on 3 Apr 2017

@dncpax it might be related to Webpack. Either way, we are backing off of Metabase for now because of it.

deitch on 8 Apr 2017

@maesenka @luc-poynt @dncpax @deitch I've opened a PR for this at #4740. It would be amazing if you could try it out and let me know how it works. It should require very little proxy configuration (I've included examples for nginx and caddy).

tlrobinson on 10 Apr 2017

@tlrobinson Hi there. I can try it out. Can you send the jar file?

dncpax on 10 Apr 2017

@dncpax Sure, here's a jar: https://www.dropbox.com/s/hyfspysmn5h50qg/metabase-pr-4740.jar?dl=1

tlrobinson on 10 Apr 2017

@jodok @TheMP @nouney @haolez @jalalmostafa @wvengen @jornh @lkraav @Jacob-Kroeze @lxchavez @moumny @VikramTiwari If you all are still interested in this feature it would be great if you could test it out. See the jar posted above and PR #4740

tlrobinson on 10 Apr 2017

can i just replace my metabase jar 0.23.1 with this one? and switch back?

dncpax on 11 Apr 2017

Hi there, never mind my question above. I installed a new instance. So far so good. Everything is working just want to try sharing and embedding.

dncpax on 11 Apr 2017

posted results on PR #4740

dncpax on 11 Apr 2017

hi @tlrobinson i noticed an issue with the download option. CSV / JSON apparently don't use the /metabase prefix and try to post directly to /api. Is this fixable?

Cannot POST /api/dataset/csv

screen shot 2017-04-14 at 3 28 22 am

aaronryden on 14 Apr 2017

👍1

@aaronryden Thanks, I've fixed that.

Please put any more issues you find in #4740

tlrobinson on 15 Apr 2017

👎1 👍1

@aaronryden @tlrobinson I have the same problem.

My CSV/ JSON/ XLSX downloads does not go with the nginx configuration. this is my conf, please let me know what wrong i am doing.

server {
listen 80 default_server;
listen [::]:80 default_server;
root /usr/share/nginx/html;

location /metabase/ {
proxy_pass http://localhost:3000/;
proxy_force_ranges on;
proxy_set_header Host $host;
proxy_set_header X-Originator-IP $remote_addr;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
proxy_set_header X-Forwarded-Proto $scheme;

    client_max_body_size 10m;
}

}

Siddarth434 on 27 Sep 2018

@tlrobinson , any solution? i am using v0.30.4

Siddarth434 on 4 Oct 2018

Hi @Siddarth434

We are runnnining metabase from a non-root app context. I just tested our nginx config with metabase 0.30.4 and it seems to be working properly even downloading CSVs works.
I am not an nginx guru so I cannot really tell you what does not work with your config but here is our config:

server { 
  server_name {{ nginx_hostname }};

  ignore_invalid_headers off; #pass through headers from metabase which are considered invalid by Nginx server.
  gzip on;
  gzip_comp_level 4;
  gzip_types text/html text/plain text/css application/json application/x-javascript text/xml application/xml application/xml+rss text/javascript;

 location /{{ app.name }}/metabase/ {
    proxy_redirect off;

    # Settings for Websocket proxying (see http://nginx.org/en/docs/http/websocket.html)
    # Based on https://github.com/metabase/metabase/blob/master/bin/aws-eb-docker/.ebextensions/metabase_config/metabase-setup.sh
    proxy_http_version  1.1;
    proxy_set_header    Upgrade         $http_upgrade;
    proxy_set_header    Connection      $connection_upgrade;
    proxy_set_header    Host            $host;
    proxy_set_header    X-Real-IP       $remote_addr;

    # Default timeout happens after 60 seconds. Increasing for long time running queries
    # Based on https://github.com/metabase/metabase/blob/master/docs/troubleshooting-guide/datawarehouse.md
    # and https://www.scalescale.com/tips/nginx/504-gateway-time-out-using-nginx/
    proxy_connect_timeout       200;
    proxy_send_timeout          200;
    proxy_read_timeout          200;
    send_timeout                200;

    proxy_pass http://localhost:{{ app.port }}/;
  }
}

Maybe that helps!

Digma on 4 Oct 2018

Same here for me. Simple nginx reverse proxy config works. Except for favicon.ico that fails since the directive does not seem to be included in the html code.

location /metabase/ {
    proxy_pass http://<machine ip>:3000/;
}

eirannejad on 7 Nov 2018

@Digma I am using v0.30.4,but don't work.Not docker,run by jar.this is my conf, please let me know what wrong i am doing.

`server {
# server listening
listen *:80;
server_name -;

ignore_invalid_headers off; #pass through headers from metabase which are considered invalid by Nginx server.
gzip on;
gzip_comp_level 4;
gzip_types text/html text/plain text/css application/json application/x-javascript text/xml application/xml application/xml+rss text/javascript;

location /dashboard/metabase/ {
proxy_redirect off;

# Settings for Websocket proxying (see http://nginx.org/en/docs/http/websocket.html)
# Based on https://github.com/metabase/metabase/blob/master/bin/aws-eb-docker/.ebextensions/metabase_config/metabase-setup.sh
proxy_http_version  1.1;
proxy_set_header    Upgrade         $http_upgrade;
proxy_set_header    Connection      $connection_upgrade;
proxy_set_header    Host            $host;
proxy_set_header    X-Real-IP       $remote_addr;

# Default timeout happens after 60 seconds. Increasing for long time running queries
# Based on https://github.com/metabase/metabase/blob/master/docs/troubleshooting-guide/datawarehouse.md
# and https://www.scalescale.com/tips/nginx/504-gateway-time-out-using-nginx/
proxy_connect_timeout       200;
proxy_send_timeout          200;
proxy_read_timeout          200;
send_timeout                200;

proxy_pass http://localhost:3005/;

}
}`

jiaweicat on 30 Nov 2018

I updated my nginx config to this and it works very well now. (Metabase is running in a docker container and exposing port# 3000). Even the /favicon.ico GET works.

location /mb/ {
    rewrite ^/mb/(.*)$ http://subdomain.company.com:3000/$1;
}

eirannejad on 30 Nov 2018

I've started collecting example configurations here: https://github.com/tlrobinson/metabase-proxy-examples, just Caddy so far because that's what I've been using lately.

If anyone wants to contribute equivalent Nginx/Apache/etc examples that would be great.

tlrobinson on 30 Nov 2018

This worked fine in v0.30.4, i.e. prior to your patch.
browser URL: https://www.example.com/reporting/

stock docker image, setup site-url and configure nginx like this:

site-url: https://www.example.com/reporting/

upstream metabase-prod {
    server {{.Address}}:{{.Port}}
}
        location /reporting {
                proxy_read_timeout 2400;
                proxy_pass http://metabase-prod;
                rewrite  ^/reporting/(.*)  /$1 break;
                proxy_set_header   Host $host;
                }

This doesn't work anymore in v0.31.2

instead I get warning messages in the console:

Warning: the Metabase site URL basename "/reporting/" does not match the actual basename "/". reporting:36:21
You probably want to update the Site URL setting to "https://www.example.com/"

Which makes little sense. Sure docker is answering requests at /, but the external browser URL is /reporting/

turning off the rewrite doesn't help because the URL's you are sending to the browser are things like this:

<link href="app/dist/vendor.bundle.css?00f24a2eeccb0afbc72602c65192b43e"

when they should be $site-url/app/dist/vendor.bundle.css?00f24a2eeccb0afbc72602c65192b43e

Setting site-url as suggested:

metabase@dev=# select value from setting where key='site-url';
                 value
---------------------------------------
 https://www.example.com/
(1 row)

returns this:

Warning: the Metabase site URL basename "//" does not match the actual basename "/". reporting:36:21
You probably want to update the Site URL setting to "https://www.example.com/"

and of course still doesn't work, to no-one's surprise.

Or do you see all this working some other way?

csawyerYumaed on 16 Jan 2019

@csawyerYumaed
I haven't tested, but I'm pretty sure it's because your rewrite is after the proxy_pass.

location /reporting {
  rewrite ^/reporting/(.*) /$1 break;
  proxy_read_timeout 2400;
  proxy_set_header Host $http_host;
  proxy_pass http://metabase-prod/;
}

flamber on 17 Jan 2019

Same here for me. Simple nginx reverse proxy config works. Except for favicon.ico that fails since the directive does not seem to be included in the html code.
location /metabase/ {
    proxy_pass http://<machine ip>:3000/;
}

Are you able to download(CSV, XLSX) the embedded reports through Metabase?

Siddarth434 on 17 Jan 2019

@csawyerYumaed
I haven't tested, but I'm pretty sure it's because your rewrite is after the proxy_pass.
location /reporting {
  rewrite ^/reporting/(.*) /$1 break;
  proxy_read_timeout 2400;
  proxy_set_header Host $http_host;
  proxy_pass http://metabase-prod;
}

It works the embed part and dashboard download part but report export to CSV Fails as the configuration does not take through the location path defined in the nginx. I am not using docker, i am running the JAR file on the server directly.

Instead of download going this way:
https:///metabase/api/public/card/

It goes this way, which results in saying Forbidden page
https:///api/public/card/

My Nginx Configuration is:

location /metabase/ {

proxy_pass http://localhost:3000/;
client_max_body_size 10m;
}

Siddarth434 on 17 Jan 2019

Hi @Siddarth434

We are runnnining metabase from a non-root app context. I just tested our nginx config with metabase 0.30.4 and it seems to be working properly even downloading CSVs works.
I am not an nginx guru so I cannot really tell you what does not work with your config but here is our config:

server { 
  server_name {{ nginx_hostname }};

  ignore_invalid_headers off; #pass through headers from metabase which are considered invalid by Nginx server.
  gzip on;
  gzip_comp_level 4;
  gzip_types text/html text/plain text/css application/json application/x-javascript text/xml application/xml application/xml+rss text/javascript;

 location /{{ app.name }}/metabase/ {
    proxy_redirect off;

    # Settings for Websocket proxying (see http://nginx.org/en/docs/http/websocket.html)
    # Based on https://github.com/metabase/metabase/blob/master/bin/aws-eb-docker/.ebextensions/metabase_config/metabase-setup.sh
    proxy_http_version  1.1;
    proxy_set_header    Upgrade         $http_upgrade;
    proxy_set_header    Connection      $connection_upgrade;
    proxy_set_header    Host            $host;
    proxy_set_header    X-Real-IP       $remote_addr;

    # Default timeout happens after 60 seconds. Increasing for long time running queries
    # Based on https://github.com/metabase/metabase/blob/master/docs/troubleshooting-guide/datawarehouse.md
    # and https://www.scalescale.com/tips/nginx/504-gateway-time-out-using-nginx/
    proxy_connect_timeout       200;
    proxy_send_timeout          200;
    proxy_read_timeout          200;
    send_timeout                200;

    proxy_pass http://localhost:{{ app.port }}/;
  }
}

Maybe that helps!

@Digma ,

My Nginx Configuration is:

location /metabase/ {

proxy_pass http://localhost:3000/;
client_max_body_size 10m;
}

Not able to download embedded reports. could you have a look if i am missing anything?

Siddarth434 on 17 Jan 2019

No difference:

        location /reporting/ {
                rewrite  ^/reporting/(.*)  /$1 break;

                proxy_read_timeout 2400;
                # Settings for Websocket proxying (see http://nginx.org/en/docs/http/websocket.html)
                # Based on https://github.com/metabase/metabase/blob/master/bin/aws-eb-docker/.ebextensions/metabase_config/metabase-setup.sh
                proxy_http_version  1.1;
                proxy_set_header    Upgrade         $http_upgrade;
                proxy_set_header    Connection      $connection_upgrade;
                proxy_set_header    Host            $host;
                proxy_set_header    X-Real-IP       $remote_addr;
                proxy_pass http://metabase-dev;

                }

Console output:

Warning: the Metabase site URL basename "/reporting/" does not match the actual basename "/". reporting:36:21
You probably want to update the Site URL setting to "https://www.example.com/"

Which is not really true. The actual basename from docker's perspective is / but the actual basename from the end-user perspective is /reporting/

csawyerYumaed on 17 Jan 2019

I'm still experiencing this issue
docker metabase v0.32.1, with nginx

# docker
docker run -d -p 3005:3000 --name metabase metabase/metabase

# nginx
location /metabase/ {
    proxy_pass http://127.0.0.1:3005;
}

Assets file still in /app/dist/ path with <base href="/" />

So where can I configure this base tag? It's never mentioned in docs

daorren on 1 Apr 2019

@daorren You need to update the "Site URL" in Metabase Admin > Settings > General.

flamber on 1 Apr 2019

@flamber Thank you for your information!

daorren on 1 Apr 2019

Is there a way to update Site URL without having the site up? Env var or cli startup param?

redchuck on 11 Apr 2019

@camsaul I would expect MB_SITE_URL to work but it doesn't seem to have any effect. I know there's some weirdness around how we store that setting due to stripping trailing slashes. How hard would it be to get MB_SITE_URL to work?

tlrobinson on 11 Apr 2019

@redchuck In the meantime the easiest way is probably to connect to the Metabase application database and update the setting. Something like:

UPDATE SETTING
SET VALUE = 'whatever'
WHERE KEY = 'site-url'

Make sure the value doesn't have a trailing slash.

tlrobinson on 11 Apr 2019

@tlrobinson MB_SITE_URL should work but since the setting has a custom magic getter that might be overriding the env var. If that's the case that's not how I meant for it to work and something I can fix with the Settings framework generally. I will investigate this and fix

camsaul on 11 Apr 2019

@redchuck it turns out MB_SITE_URL actually does work, if you include the trailing slash. See the last couple comments in #9764.

tlrobinson on 17 Apr 2019

👍1

For me it does not load correctly i have used the mb_site_url to define a address hostname:8051/metabase it seems like the data is not downloading completly so the title displays metabase and some content can be loaded but not all.

@tlrobinson is this the correct way?
Define a server for the metabase port
upstream metabase-backend { server metabase:3000; }
location /metabase/{ proxy_pass http://metabase-backend; }
Response in chrome

in gui

The version is docker metabase v0.32.9

ZerNox on 3 Jul 2019

@ZerNox I don't have time to look into this right now, but I've got some example configurations for Caddy available at https://github.com/tlrobinson/metabase-proxy-examples (accepting PRs for Nginx examples!)

tlrobinson on 3 Jul 2019

Still having this issue with MB_SITE_URL. the browser tried access to domain.tld/app/dist/...js first with 404 and then tried domain.tld/metabase/app/dist/...js. The latter path is correct, however blocked by both Firefox and Chrome.

Firefox complains differently, but by large the same

The resource from “http://domain.tld/metabase/app/dist/app-main.bundle.js?0d1325582067a4b57db2” was blocked due to MIME type (“text/html”) mismatch (X-Content-Type-Options: nosniff).

I'm using v0.32.9 in docker with nginx reverse proxy from the host.

Congee on 17 Jul 2019

👍3

The issue still persists in v0.33.4. Metabase doesn't reacts to the changes in in Site Url even from GUI, database, env. Works from root only. Is it going to be fixed? Thanks.

amaris12 on 24 Oct 2019

@amaris12 There's a regression in 0.33.4, which will be fixed by #11206, and most likely be in 0.33.5

flamber on 24 Oct 2019

Metabase: Allow Metabase to run at a non-root app context such as /mb/

Most helpful comment

All 64 comments

Related issues