Unable to enable caching of GetMap/GetTile requests

[omad]

Description

I'm trying to setup CloudFront (AWS's CDN/cache) to work in front of a datacube-ows deployment.

We have quite a complex configuration, but I believe I've setup the correct options. GetCapabilities requests are now being cached, but GetMap/GetTile requests are not, and I can't work out why.

I'm expecting cache-control: max-age 3600 headers to be added to my requests, but instead they're still getting cache-control: no-cache headers.

Steps to Reproduce

curl -I 'https://ows.dea.ga.gov.au/?transparent=true&format=image%2Fpng&exceptions=XML&styles=true_colour&version=1.3.0&time=2023-01-24&service=WMS&request=GetMap&layers=ga_ls_ard_3&bbox=15536896.1173580
62%2C-2113330.958028555%2C15576031.875840079%2C-2074195.199546542&width=256&height=256&crs=EPSG%3A3857'
HTTP/2 200 
content-type: image/png
content-length: 167321
date: Wed, 25 Jan 2023 04:34:42 GMT
server: gunicorn
access-control-allow-origin: *
cache-control: no-cache
x-request-id: 1-63d0b161-57cc35d054da33cb162ca65c
x-cache: Miss from cloudfront
via: 1.1 5e473a5e64c6a2f7bc916721cc188252.cloudfront.net (CloudFront)
x-amz-cf-pop: SYD62-P2
x-amz-cf-id: WV69MgOnjdDM1i5QkKVfd_HjZiSb_4zHw8qaoRVDu7UdKSnLP8JIjg==

I've checked as best I can that the expected configuration is being loaded:

<<K9s-Shell>> Pod: web/ows-datacube-ows-prod-b47db98-7fjdd | Container: ows 
ows@ows-datacube-ows-prod-b47db98-7fjdd:~$ python
Python 3.10.6 (main, Nov 14 2022, 16:10:14) [GCC 11.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> from datacube_ows.startup_utils import parse_config_file
>>> cfg = parse_config_file()
>>> cfg.layers[0].child_layers[0].resource_limits.wms_cache_rules
<datacube_ows.resource_limits.CacheControlRules object at 0x7fab72b34f70>
>>> cfg.layers[0].child_layers[0].resource_limits.wms_cache_rules.rules
[{'min_datasets': 1, 'max_age': 3600}, {'min_datasets': 5, 'max_age': 86400}, {'min_datasets': 9, 'max_age': 1209600}]
>>> cfg.layers[0].child_layers[0].name
'ga_ls_ard_3'

But when I try using this CacheControlRules object, it's not behaving as I'd expect.

>>> cfg.layers[0].child_layers[0].resource_limits.wms_cache_rules.cache_headers(1)
{'cache-control': 'no-cache'}
>>> cfg.layers[0].child_layers[0].resource_limits.wms_cache_rules.cache_headers(2)
{'cache-control': 'no-cache'}
>>> cfg.layers[0].child_layers[0].resource_limits.wms_cache_rules.cache_headers(3)
{'cache-control': 'no-cache'}
>>> cfg.layers[0].child_layers[0].resource_limits.wms_cache_rules.cache_headers(4)
{'cache-control': 'no-cache'}

Argh, I've solved it, it seems that max_datasets hasn't been specified, and is defaulting to 0, which is disabling caching!

I can fix it by setting it, but, it also seems like not a good default.

>>> cfg.layers[0].child_layers[0].resource_limits.wms_cache_rules.max_datasets = 20
>>> cfg.layers[0].child_layers[0].resource_limits.wms_cache_rules.cache_headers(6)
{'cache-control': 'max-age=86400'}

Context (Environment)

datacube-ows version (datacube-ows --version):

ows@ows-datacube-ows-prod-b47db98-7fjdd:~$ datacube-ows --version
 * Serving Flask app 'datacube_ows'
 * Debug mode: off
[2023-01-25 04:43:58,592] [INFO] WARNING: This is a development server. Do not use it in a production deployment. Use a production WSGI server instead.
 * Running on http://127.0.0.1:5000
[2023-01-25 04:43:58,592] [INFO] Press CTRL+C to quit

Whoops, looks like --version isn't implemented?

ows_config.py file (link, sample code)

https://github.com/GeoscienceAustralia/dea-config/blob/master/prod/services/wms/ows_refactored/ows_reslim_cfg.py#L4-L17

datacube product metadata (datacube product show product_name)

[SpacemanPaul]

Agreed this is a bug - WMS caching should be possible without a WMS max_datasets being set.