← Back to context

Comment by fidotron

2 days ago

It's completely nuts that Firebase has this: https://status.firebase.google.com/incidents/ZcF1YDUvpdixZ2e...

"Firebase Data Connect unavailable due to a known Google Cloud global outage"

While the Google Cloud status page https://status.cloud.google.com/ says "No major incidents" and everything is green. So Google Cloud know there is an outage but just deem it not major enough to show it.

Edit to add: within 10 minutes of this post Google updated their status page. More curiously the Firebase page I linked to has been edited to remove mention of Google Cloud in the status and now says "Firebase Data Connect is currently experiencing a service disruption. Please check back for status. ".

IIRC status pages drive customer compensation for downtime. Updating it is basically signing the check for their biggest customers, in most similar companies you need a very senior executive to approve the update

On the other side of this, Firebase probably doesn't have money at stake making the update

  • It is not the status page that drives customer compensation. It is downtime.

    • The status page is essentially an admission of guilt. It can require approval from the legal department and a high level official from the company to approve updating it and the verbiage used on the status page.

      10 replies →

  • Nah, its just some client side caching / JS stuff. Clicking the big refresh button fixed it for me, 15 minutes before OP noted it.

    (n.b. as much as Google in aggregate is evil, they're smart evil. You can't avoid execs approving every outage because checks without some paper trail, and execs don't want to approve every outage, you'd have to rely on too many engineers and sales people, even as ex-employees, to keep it a secret. disclaimer: xoogler)

    (EDIT: for posterity, we're discussing a "overall status" thing with a huge refresh button, right above a huge table chockful of orange triangles that indicate "One or more regions affected" - even when the "overall status" was green, the table was still full of orange and visible immediately underneath. My point being, you gotta suppose a wholeeee bunch of stuff to get to the point there was ever info suppressed, much less suppressed intentionally to avoid cutting checks)

Something must be preventing them updating the status page at this point. Of course they could still deem it not enough, but just from my limited tests, docker, buf, etc (it may not be GCP that is down, but it is quite the coincidence). are outright down. I'd wager that this is much more widespread.

This extra funny that GCP status page even includes a “last updated” time, which is exactly built to convey possible failure to update in cases like this

No major incident as of “ Last updated time: 12 Jun 2025, 11:48 PDT”

Maybe their dashboard is hosted on GCP and they are displaying a cached version. :-)

More likely they are unable to update their own status page, but in either case not covering themselves in glory over at GCP right now.

AWS has this all the time. If you need to know if a service is down in a region, check for other engineers talking about it on X.