Google Meet was hit by an outage that made it difficult to initiate calls, except for a few frustrated customers who were greeted with blank screens and “Joining…” spinners and unable to accomplish much else. The company recognized the issue on its Workspace Status Dashboard, listing problems that had impacted loading the Meet interface and joining meetings, before noting that engineers had addressed the issue by undoing a recent change.
What Google said
In a status notice, Google said that some Meet customers faced slow load times or couldn’t join meetings. The incident was the result of a configuration change in its content edge caching layer — the distributed system it uses for rapidly delivering web content — and service stability was restored after the change was reversed, the company said. Everybody was glued to their recovery, as traffic settled.
While Google doesn’t suggest any security or data-loss heartache, it’s possible that the symptoms implied a client-side initialization failure, and not a core backend explosion, as things seem grossly unaffected at the back-end if not for the browsers. In simple terms: the foundation of the call infrastructure appears to have remained accessible, but the components needed to run and initiate the meeting in the browser were at times unreachable.
What users experienced
Reports collated on the crowd-sourced web incident tracker Downdetector at the height of the outage clocked in at more than 312,000 as the outrage unfolded, featuring users in clusters across North America, Europe, and large swaths of Asia. On social media, users reported delays in participants loading in the meeting UI, error messages when trying to join from Calendar links, and calls that dropped back to the lobby.
Anecdotally, many users commented that switching to a different device or moving from the desktop browser to the mobile app sometimes resolved the issue, further suggesting that the issue was not with Meet’s core conferencing services but with how web assets were being served at the edge.
Why a simple cache change can disrupt video calls
Modern collaboration apps rely on content delivery networks that live close to users and quickly deliver static assets such as JavaScript bundles, styles, and localization files. If a bad config hits those edge nodes, browsers won’t be able to process the essential assets, making the (eg) service web half-loaded or broken despite the fact his application’s servers are healthy.
Google’s site reliability engineering playbook is big on the chestnut of quick rollback when a recent change is associated with a networked fault. Reverting from one of these changes simply puts the last known # working configuration back in place, and invalidates incorrect data, however, full recovery may take several minutes to propagate through the world-wide network of caches as each cache will take its own time to expire old data and fetch new data.
Who felt the impact
Google Meet is part of Google Workspace, which is used by billions of people and millions of businesses — from small businesses to schools to hospitals and more. Even brief outages can derail interviews, telehealth check-ins and classes. Research from the industry group Uptime Institute has shown that the business cost of outages is increasing, with a larger percentage now costing six figures [for] every occurrence, so catching and reversing errors matters.
It was the timing and its dispersion across the map for disruption that made the impact in some places pronounced: When outages hit during common meeting hours in region after region, a single missed standup or client briefing can bloom into rescheduled meetings and lost productivity for teams that depend on regular calls.
Workarounds to try with Meet issues
When Meet fails, there are a couple of strategies to keep a conversation on track while engineers work the problem:
- Verify scope using the Google Workspace Status Dashboard. If Meet is impacted, anticipate recovery updates as fixes are introduced.
- Go an another way: the mobile app; the Progressive Web App; a new browser profile. You can clear cache or use an incognito window to avoid trouble from cached assets.
- Connect by phone using dial-in numbers included with various editions of Workspace if the UI won’t load. Calendar invites usually contain local numbers and a PIN.
- Share materials asynchronously: Share slides or notes through Docs and record a short update through some other approved method while waiting for service to recover.
The broader picture for reliability
Edge configuration-change incidents are a familiar trade-off in today’s highly-distributed world: the same mechanisms that bring speed at scale also widen the blast radius of a bad push. The good news is that rollbacks are well-practiced and, done fast enough, can minimize downtime.
For customers that rely on Meet, the recipe for this redundancy looks like: add backup dial-in numbers by default, document workarounds for mission-critical calls and give admin the ability to deliver updates rapidly. As today’s hitch demonstrated, even the biggest platforms can hiccup — but clear communication and preparations for when things go wrong turn a canceled meeting into a speed bump instead of a day-derailing disaster.