collections - speed

[danijelk]

Hi,

we've used statamic for some time now, and after spending time upgrading to 3.3 we now want to try to find out how we can speed up local development as pagespeed can be quite slow. We've created a unittest which should crawl all pages to assert status 200, which isn't going all too well (timeouts and freezes in CI), and we wanted to share our findings and thoughts on where there is room for improvements.

background
It's a fairly simple site (https://vemco.group) with multi lang support enabled.

cache

• helps to go from the default 5-15s to 170ms. But gives unexpected behaviour as it seems to cache 'too' good.
  CACHE_DRIVER=array
  CACHE_STRATEGY=full

• gets it below 100ms, best environment so far
  CACHE_DRIVER=redis
  CACHE_STRATEGY=half

• slow default 5-15s speed
  CACHE_DRIVER=redis
  CACHE_STRATEGY=null

xdebug
disabling xdebug shaves off a couple of seconds, we run a fairly heavy analytics project with thousands of models loaded on pages, without this slowness, but we can't point out why xdebug would cause statamic to go so slow.

unittest
We simply get all routes and try to simply visit it with phpunit, but it fails when it visits some specific urls it seems for now

$ php artisan statamic:stache:warm
Please wait. This may take a while if you have a lot of content.
Warming Stache...
You have poured oil over the Stache and polished it until it shines. It is warm and ready
$ $CI_PROJECT_DIR/vendor/bin/phpunit tests
PHPUnit 9.5.21 #StandWithUkraine
[2022-07-08 13:12:16] local.INFO: /segments/banking-services  
[2022-07-08 13:12:17] local.INFO: /resources/blog  
[2022-07-08 13:12:21] local.INFO: /resources/cases  
[2022-07-08 13:12:22] local.INFO: /about-us/contact  
[2022-07-08 13:12:23] local.INFO: /cookie-policy  
[2022-07-08 13:12:24] local.INFO: /solution/covid-19  
[2022-07-08 13:12:25] local.INFO: /about-us/clients  
[2022-07-08 13:12:26] local.INFO: /resources/downloads
... snip
[2022-07-08 13:12:58] local.INFO: /segments/retailers-malls  
[2022-07-08 13:13:14] local.INFO: /solution/sales-analytics  
[2022-07-08 13:13:50] local.INFO: /segments/schools-universities-libraries  
Killed
Cleaning up project directory and file based variables
00:02
ERROR: Job failed: exit code 137

-- stuck and timeout after 8min

collections
Stripping everything away from templates didn't help much, what does help is when we move our ~60 blogs out of the collections folder. It's about a 1-2sec increase per language that we move out.
After moving all blogs out we're down to 1s loading time (with the default slowness cache setup) which is the acceptable timing.

final
So our focus are on the collections

• it seems like they are loaded on each page, wether they are used or not - might be worth looking into lazy-loading them instead?
• collections for all languages are loaded, even the language that is not selected, and even languages that aren't enabled - should be fairly easy to ignore folders that do not need to be loaded
• making a new cache variable for STACHE to cache collections only and avoid the full cache_strategy as this can cause unexpected behaviour. Not sure what 'half' strategy really does?

[jasonvarga]

Have you tried disabling the watcher in your tests? If you've got lots of files, the watcher will be traversing them every test and slow things down unnecessarily.

https://github.com/statamic/statamic/blob/da2f408fb219534b707481b04c33c4e6acea0fe0/config/statamic/stache.php#L16

https://statamic.dev/stache#watcher

[danijelk]

Hiya Jason,

Yes tried it before as well - forgot to mention it, not much changes with it off in our project.
But I have found the CI blocks were memory related, so might have hit a memory leak in phpunit.
Refactored the test to use Http::async() with artisan serve, and it goes through it all without any issues in 4+min.

Could you maybe point out where the collections are scanned/handled, would like to spend some time understanding the logics.

[2022-07-08 14:36:04] testing.INFO: /segments/banking-services
[2022-07-08 14:36:04] testing.INFO: memory_get_usage: 31.846296
[2022-07-08 14:36:04] testing.INFO: /resources/blog
[2022-07-08 14:36:04] testing.INFO: memory_get_usage: 33.350088
[2022-07-08 14:36:04] testing.INFO: /resources/cases
[2022-07-08 14:36:04] testing.INFO: memory_get_usage: 33.42416
[2022-07-08 14:36:04] testing.INFO: /about-us/contact
[2022-07-08 14:36:04] testing.INFO: memory_get_usage: 33.31876
[2022-07-08 14:36:04] testing.INFO: /cookie-policy
[2022-07-08 14:36:04] testing.INFO: memory_get_usage: 33.333288
[2022-07-08 14:36:04] testing.INFO: /solution/covid-19
[2022-07-08 14:36:04] testing.INFO: memory_get_usage: 33.3256
[2022-07-08 14:36:04] testing.INFO: /about-us/clients
[2022-07-08 14:36:04] testing.INFO: memory_get_usage: 33.366832
[2022-07-08 14:36:04] testing.INFO: /resources/downloads
[2022-07-08 14:36:04] testing.INFO: memory_get_usage: 33.3648
[2022-07-08 14:36:04] testing.INFO: /data-analytics-in-us-canada
[2022-07-08 14:36:04] testing.INFO: memory_get_usage: 33.419224
[2022-07-08 14:36:04] testing.INFO: /solution/facilities-management
[2022-07-08 14:36:04] testing.INFO: memory_get_usage: 33.4814
[2022-07-08 14:36:04] testing.INFO: /
[2022-07-08 14:36:04] testing.INFO: memory_get_usage: 33.660144
[2022-07-08 14:36:04] testing.INFO: /home_demo
[2022-07-08 14:36:04] testing.INFO: memory_get_usage: 34.019872
[2022-07-08 14:36:04] testing.INFO: /segments/hotels-entertainment-services
[2022-07-08 14:36:04] testing.INFO: memory_get_usage: 34.684208
[2022-07-08 14:36:04] testing.INFO: /resources/news
[2022-07-08 14:36:04] testing.INFO: memory_get_usage: 36.11824
[2022-07-08 14:36:04] testing.INFO: /solution/occupancy-management
[2022-07-08 14:36:04] testing.INFO: memory_get_usage: 39.030112
[2022-07-08 14:36:04] testing.INFO: /solution/pedestrian-counting
[2022-07-08 14:36:04] testing.INFO: memory_get_usage: 45.061064
[2022-07-08 14:36:04] testing.INFO: /solution/people-counting
[2022-07-08 14:36:04] testing.INFO: memory_get_usage: 57.34988
[2022-07-08 14:36:04] testing.INFO: /privacy-policy
[2022-07-08 14:36:04] testing.INFO: memory_get_usage: 82.487192
[2022-07-08 14:36:05] testing.INFO: /solution/queue-management
[2022-07-08 14:36:05] testing.INFO: memory_get_usage: 133.760312
[2022-07-08 14:36:05] testing.INFO: /about-us/resellers
[2022-07-08 14:36:05] testing.INFO: memory_get_usage: 238.564976
[2022-07-08 14:36:06] testing.INFO: /segments/retailers-malls
[2022-07-08 14:36:06] testing.INFO: memory_get_usage: 452.33364
[2022-07-08 14:36:08] testing.INFO: /solution/sales-analytics
[2022-07-08 14:36:08] testing.INFO: memory_get_usage: 888.465176