Cloudflare Developers•7mo ago

Workers fetch cache (tiered caching) - How to handle conditional caching?

I have a use-case where we'd like to utilize tiered caching instead of the cache API to help shield our origins from requests, but I'm struggling to replicate the same logic using the fetch cache. Our worker essentially acts as a reverse proxy, handling internal routing & query parameters to the origins (multiple servers). This is a very slimmed down version of our logic using the cache API:

export async function getReverseProxyResponseCacheApi(request: Request, requestRules: RequestRules, pageProviderUrl: string): Promise<Response> {
    // Disable caching when requesting on an internal IP
    const useCache = !SecurityService.isBusinessIpRequest(request);

    const cache = caches.default;

    // Appends various information to the URL to ensure that the cache key is unique. We include internal data
    // such as experiments, device, region, currency info, etc.
    const cacheKey = CacheKeyService.getCacheKey(request, pageProviderUrl);

    let page = useCache ? await cache.match(cacheKey) : undefined;

    if (!page) {
        const pageProviderResponse = await fetch(pageProviderUrl);
        page = new Response(pageProviderResponse.body, pageProviderResponse);

        if (useCache && page.ok) {
            const cacheControl = pageProviderResponse.headers.get('cache-control');
            // If the cache is not public, or set to no-store, don't cache
            if (cacheControl && !cacheControl.includes('no-store') && cacheControl.includes('public')) {
                page.headers['Cache-Tag'] = CacheKeyService.getCacheTags(request, requestRules);

                // Respects the cache-control header TTL that was present on the origin response
                await cache.put(cacheKey, page);
            }
        }
    }

    // If the page still isn't found, throw an error
    if (!page) {
        throw new Error('Page not found');
    }

    return page;
}

export async function getReverseProxyResponseCacheApi(request: Request, requestRules: RequestRules, pageProviderUrl: string): Promise<Response> {
    // Disable caching when requesting on an internal IP
    const useCache = !SecurityService.isBusinessIpRequest(request);

    const cache = caches.default;

    // Appends various information to the URL to ensure that the cache key is unique. We include internal data
    // such as experiments, device, region, currency info, etc.
    const cacheKey = CacheKeyService.getCacheKey(request, pageProviderUrl);

    let page = useCache ? await cache.match(cacheKey) : undefined;

    if (!page) {
        const pageProviderResponse = await fetch(pageProviderUrl);
        page = new Response(pageProviderResponse.body, pageProviderResponse);

        if (useCache && page.ok) {
            const cacheControl = pageProviderResponse.headers.get('cache-control');
            // If the cache is not public, or set to no-store, don't cache
            if (cacheControl && !cacheControl.includes('no-store') && cacheControl.includes('public')) {
                page.headers['Cache-Tag'] = CacheKeyService.getCacheTags(request, requestRules);

                // Respects the cache-control header TTL that was present on the origin response
                await cache.put(cacheKey, page);
            }
        }
    }

    // If the page still isn't found, throw an error
    if (!page) {
        throw new Error('Page not found');
    }

    return page;
}

However, with the fetch cache, we don't have the same control as far as I can tell.

export async function getReverseProxyResponseFetchCacheApi(request: Request, requestRules: RequestRules, pageProviderUrl: string): Promise<Response> {
    const useCache = SecurityService.isBusinessIpRequest(request);
    const cacheKey = CacheKeyService.getCacheKey(request, pageProviderUrl);

    let page = await fetch(pageProviderUrl, {
        cf: {
            cacheTtl: useCache ? requestRules.reverseProxyTtl : undefined,
            cacheTtlByStatus: {
                '300-599': 0,
            },
            cacheKey: cacheKey,
            cacheTags: [CacheKeyService.getCacheTags(request, requestRules)],
        },
    });

    page = new Response(page.body, page);

    // Cache-Control is completely ignored by the tiered cache since we provided a TTL
    const cacheControl = page.headers.get('cache-control');

    // If the cache is not public, or set to no-store, don't cache
    if (!page.ok || (cacheControl && (!cacheControl.includes('public') || cacheControl.includes('no-store')))) {
        // How do we handle this?
        // await caches.default.delete(cacheKey);
        return page;
    }

    return page;
}

export async function getReverseProxyResponseFetchCacheApi(request: Request, requestRules: RequestRules, pageProviderUrl: string): Promise<Response> {
    const useCache = SecurityService.isBusinessIpRequest(request);
    const cacheKey = CacheKeyService.getCacheKey(request, pageProviderUrl);

    let page = await fetch(pageProviderUrl, {
        cf: {
            cacheTtl: useCache ? requestRules.reverseProxyTtl : undefined,
            cacheTtlByStatus: {
                '300-599': 0,
            },
            cacheKey: cacheKey,
            cacheTags: [CacheKeyService.getCacheTags(request, requestRules)],
        },
    });

    page = new Response(page.body, page);

    // Cache-Control is completely ignored by the tiered cache since we provided a TTL
    const cacheControl = page.headers.get('cache-control');

    // If the cache is not public, or set to no-store, don't cache
    if (!page.ok || (cacheControl && (!cacheControl.includes('public') || cacheControl.includes('no-store')))) {
        // How do we handle this?
        // await caches.default.delete(cacheKey);
        return page;
    }

    return page;
}

Is there any way to work around this that doesn't involve using the cache purge API from within the worker? (250k purges per day would quickly be hit)

1 Reply

Cole-kun•7mo ago

AFAIK, you will need to use both Workers and your origin to use fetch cache correctly. caches.default.delete is a Cache API operation, and it is local to where the Worker ran. It won't work with Tiered Cache/Fetch Cache. One way to do it:

cf: {
  cacheEverything: true,
  cacheKey: cacheKey,
  cacheTags: do-your-stuff-etc
}

cf: {
  cacheEverything: true,
  cacheKey: cacheKey,
  cacheTags: do-your-stuff-etc
}

And then make your Origin send the necessary headers to make this work, for example, Cloudflare-CDN-Cache-Control. Fetch should respect it, not caching private content, etc.

Gaming

Programming

Workers fetch cache (tiered caching) - How to handle conditional caching?