Brief period of unavailability when publishing Workers Sites

Greetings!

We’re using Workers Sites for hosting simple status page with announcements regarding system availability while performing maintenance. We’re probably better off with specialized SaaS solution but for now WS seems like a perfect match.

I noticed somewhat surprising behavior after publishing new version of the site—when I did wrangler publish as always and hit refresh in the browser I’ve got a ‘Resource with key <…> is unavailable’ error message instead of index page.

What I think is happening is that worker script itself is not changed when we only tweak static html content. So the script remains the same and don’t get updated, but KV objects are—and due to the fact that WS stores it with some kind of hash in the name means that they will not be available until they propagate fully.

Is there a recommended way to work around that so there is no downtime upon publish?

Hi @maxgashkov,

I can confirm that this is a bug – thank you for reporting it!

Unfortunately, there’s no workaround at the moment. :frowning: I realize this is pretty bad, and basically kills your use case of using Workers Sites as a status page.

I don’t know exactly when this will be fixed, but we’re tracking the issue on Wrangler’s public GitHub repo here: https://github.com/cloudflare/wrangler/issues/783. I’d encourage you to subscribe to it.

Harris

4 Likes

Well noticed :clap:
As your objective is status pages, you might have more efficiency :moneybag: building a worker yourself.

Status pages tend to be simple and objective, and very small.
A worker that stores status on a JSON renders quicker, in this case, zero downtime, just a small 10-30 seconds lag.

1 Like