Question: is it possible to create NFS volumes with lando?

[mmenozzi]

Hi guys,
I would like to create NFS volumes like explained here: https://medium.com/@sean.handley/how-to-set-up-docker-for-mac-with-native-nfs-145151458adc

Is it possible with lando? I cannot find in the docs how to specify advanced volumes configuration with the .lando.yml.

[mmenozzi]

I think I found a solution for this. My .lando.yml looks like this:

name: myapp
recipe: lamp
config:
  webroot: .

services:
  appserver:
    overrides:
      services:
        volumes:
          - "nfsmount:/app"

compose:
  - compose.yml

And my compose.yml looks like:

version: '3'
volumes:
  nfsmount:
    driver: local
    driver_opts:
      type: nfs
      o: addr=host.docker.internal,rw,nolock,hard,nointr,nfsvers=3
      device: ":${LANDO_APP_ROOT_BIND}"

It seems to work.
Do you see any issue with such configuration?

Thank you for your feedback!

[dustinleblanc]

@mmenavas I'd only say that you are in ymmv territory, as you're overriding our standard volume mounts (which are going to be using the defaults for the platform in question), other than that, no specific issues.

I'd be interested in how your performance goes vs the standard volumes though! I am going to close this as you seem to have found a path, but do share your results if it turns out to be a boon for performance!

[mmenozzi]

Thank you for your feedback. With this configuration I have faster response time of my application but I didn’t run any benchmark of disk I/O performance. Do you have any suggestions on what tool to use to perform such benchmark?

[dustinleblanc]

Hey @mmenavas even just time curl -I http://mysite.lndo.site with both methods is probably useful. Same thing when running composer commands, drush commands, etc

[chill-genomeweb]

@mmenozzi , did you ever get a chance to benchmark using NFS (vs osxfs, I assume)? If so, I'd be really interested, even if the results are just approximate.
Thanks!

[mmenozzi]

Hi @chill-genomeweb no I didn't because I don't know how to do a reliable benchmark. Do you have any tool/idea to suggest?

[chill-genomeweb]

@mmenozzi , I'd keep it simple. You could use cURL's recording feature, but an even simpler approach is using time command line tool. Examples:

• Times the request to build the homepage of a local site: $ time curl http://mysite.lndo.site/ > /dev/null
• Flushes caches on a Drupal 7 site, but hides all output except the timing: $ time drush cc all > /dev/null

When using cURL, it's probably best to disable page caching on the website.
I'm just curious in some rough comparisons, to know whether I should try the NFS approach. Hope this helps!

[mmenozzi]

Hi, I just did a simple benchmark using Apache Benchmark (ab) with 10, not concurrent, requests to my site (a Magento 1.9 site with some customizations).

Here are the results with the lando configuration using the NFS approach:

$ ab -n 10 https://forbicy.lndo.site/it_it/
This is ApacheBench, Version 2.3 <$Revision: 1807734 $>
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Licensed to The Apache Software Foundation, http://www.apache.org/

Benchmarking forbicy.lndo.site (be patient).....done


Server Software:        Apache/2.4.10
Server Hostname:        forbicy.lndo.site
Server Port:            443
SSL/TLS Protocol:       TLSv1.2,ECDHE-RSA-AES128-GCM-SHA256,2048,128
TLS Server Name:        forbicy.lndo.site

Document Path:          /it_it/
Document Length:        139143 bytes

Concurrency Level:      1
Time taken for tests:   11.309 seconds
Complete requests:      10
Failed requests:        9
   (Connect: 0, Receive: 0, Length: 9, Exceptions: 0)
Total transferred:      1398945 bytes
HTML transferred:       1391335 bytes
Requests per second:    0.88 [#/sec] (mean)
Time per request:       1130.943 [ms] (mean)
Time per request:       1130.943 [ms] (mean, across all concurrent requests)
Transfer rate:          120.80 [Kbytes/sec] received

Connection Times (ms)
              min  mean[+/-sd] median   max
Connect:       17   37  59.4     18     206
Processing:   957 1094 216.7   1042    1690
Waiting:      953 1090 216.7   1038    1686
Total:        975 1131 274.4   1061    1895

Percentage of the requests served within a certain time (ms)
  50%   1061
  66%   1089
  75%   1094
  80%   1156
  90%   1895
  95%   1895
  98%   1895
  99%   1895
 100%   1895 (longest request)

Then I did the same with the standard Lando volume (which is a Docker bind mount in delegated mode). Here the results:

$ ab -n 10 https://forbicy.lndo.site/it_it/
This is ApacheBench, Version 2.3 <$Revision: 1807734 $>
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Licensed to The Apache Software Foundation, http://www.apache.org/

Benchmarking forbicy.lndo.site (be patient).....done


Server Software:        Apache/2.4.10
Server Hostname:        forbicy.lndo.site
Server Port:            443
SSL/TLS Protocol:       TLSv1.2,ECDHE-RSA-AES128-GCM-SHA256,2048,128
TLS Server Name:        forbicy.lndo.site

Document Path:          /it_it/
Document Length:        139159 bytes

Concurrency Level:      1
Time taken for tests:   38.344 seconds
Complete requests:      10
Failed requests:        9
   (Connect: 0, Receive: 0, Length: 9, Exceptions: 0)
Total transferred:      1399160 bytes
HTML transferred:       1391550 bytes
Requests per second:    0.26 [#/sec] (mean)
Time per request:       3834.397 [ms] (mean)
Time per request:       3834.397 [ms] (mean, across all concurrent requests)
Transfer rate:          35.63 [Kbytes/sec] received

Connection Times (ms)
              min  mean[+/-sd] median   max
Connect:       17   25  19.0     19      79
Processing:  3529 3809 256.2   3803    4302
Waiting:     3524 3804 256.0   3796    4297
Total:       3547 3834 270.8   3822    4381

Percentage of the requests served within a certain time (ms)
  50%   3822
  66%   3833
  75%   3945
  80%   4186
  90%   4381
  95%   4381
  98%   4381
  99%   4381
 100%   4381 (longest request)

The result is 3834.397 ms with Docker bind mount against 1130.943 ms with NFS mount.
Is this enough for you @chill-genomeweb and @dustinleblanc?

[chill-genomeweb]

@mmenozzi , yes, this helps a lot. Definitely want to try out this NFS approach soon.
Thank you for doing this!

[mmenozzi]

You're welcome @chill-genomeweb.
Note also that the approach described above doesn't work anymore with version beta.46 because of this commit b2373cf.

As a workaround you can use the following volumes definition for your service:

services:
  appserver:
    overrides:
      services:
        volumes:
          - "${LANDO_VOLUME}:/app"

But you have to export the LANDO_VOLUME env variable to the value nfsmount before launching the lando start command.

[dustinleblanc]

Yeah, NFS seems to be like, 3x faster, if it were a easy to implement reliably solution, maybe we'd want to look more at that

[tylers-username]

@mmenozzi - how are automating the mapping of your user & group IDs from the host to the container in any way?

[mmenozzi]

@tylerssn I don't. But I can confirm that the application running inside the container is able to write on the filesystem and a file written by the host is readable from the application as well. I think this is possible due to the -mapall option of the /etc/exports configuration.

[tylers-username]

I can write to files, however, my session save handler checks the Lando user (www-data) ID and compares it with the file owner ID which matches my host user ID. This results in a warning.

Though this is docker/nfs related, it is not Lando related so I will stop congesting this thread.

Thanks for the feedback @mmenozzi !