Hacker News Reader

Comment by chimerasaurus | Hacker News Reader

Disclaimer - I’m a Googler.

As a counterpoint, Cloud has acquired several companies that haven’t disappeared; rather, are fairly foundational in the portfolio:

- Cask (now Cloud Data Fusion as a product, CDAP as an OSS project)

- Looker

- Stackdriver

View on HN

Stackdriver did not disappear, but having used it daily for the past 4 years and witnessed its painfully slow and buggy migration into the GCP admin, I can't say it's a product that has visibly gotten much love. We kept using it for some things out of inertia, but I can't say it ever impressed me.

How so? I've only ever used it on GC, not pre-acquisition.

Prior to the acquisition Stackdriver was never best of breed among its peer group but it was acceptable. Post-acquisition the feature set appears as an end user to have more or less frozen and their UIs feel haphazard: dashboarding is a mess, the trace UI is very complex without really giving me greater query power over alternatives and the options for converting one kind of data stream into another (logs to metrics, traces to logs to metrics etc) just aren't there. Google tightly integrates Stackdriver into its other products -- I assume that's where most of the labor for the Stackdriver team has gone -- but it feels cheap and like they've had to intentionally limit options to force buy-in. For what it is, Stackdriver is surprisingly expensive and their SDKs are quite complicated for what they end up accomplishing. Why should the end-user need to internalize Stackdriver Metrics' API to be able to produce metrics about their application? (I am sympathetic to the notion that this SDK is low-level and meant to be consumed by higher-level APIs like OpenTelemetry et al).

Nowadays the Stackdriver suite just feels left behind. There's plenty of other options in this space that have a nice internal integration and are reasonably inexpensive to purchase, or run yourself. If Stackdriver weren't so tightly integrated with Google's other cloud products I have to imagine it would get even less use than it does now.

Most of the things that bothered me were fairly minor, but taken together they just didn't give me the impression that there was a lot of effort going into the product. Lots of janky UI (filtering/browsing logs being particularly cumbersome), and a bunch of things that seem promising at first but are very much "lite" versions of more powerful products like Pingdom/Sentry/DataDog. Also I always felt like the monitoring metrics were not as well integrated into the GCP system as they should have been, given that they're under the same roof...we spun up a Prometheus cluster to do it by hand instead.

Also, until recently the admin lived outside of the main GCP admin and logging into it required some kind of redirect-bouncing oauth handshake that was consistently buggy for me. Something to do with browser defaults/3rd party cookies along the chain of redirects that would cause an infinite loop. I managed to work around it on my work computer, but it was broken for at least a year...not awesome for a tool you'll be logging into in emergencies, often from other devices.

Using Stackdriver as an example here shows how out of touch people at Google are with their customers. Stackdriver has been terrible for years and has visibly become worse over time rather than better.

Firebase?

Isn't Looker a talent acquisition for Data Studio?

No, they are quite separate.

Reply on news.ycombinator.com