From duct tape to declarative: Playtika’s platform engineering journey with TeraSky and Spectro Cloud

When you’ve spent more than a decade building and running a massive Kubernetes estate across thousands of servers and multiple data centers, it’s easy to believe no one else’s playbook applies to you.

That was Playtika’s mindset for years — until a platform overhaul project led by systems integrator TeraSky and including Spectro Cloud Palette proved that even the most advanced engineering organizations can benefit from a buy-over-build mindset.

Presented at KubeCon + CloudNativeCon North America’s Platform Engineering Day under the title “From duct tape to declarative: Playtika’s platform overhaul at unicorn scale,” the talk from Scott Rosenberg, lead architect at TeraSky, and Shemer Mashiach, Director of IT Platforms and Solutions at Playtika, struck a chord with the audience. It captured the pain and payoff of evolving from a homegrown Kubernetes platform to a standardized, declarative, and sustainable one.

The early days: when everything was homegrown

Playtika, one of the world’s leading gaming companies, began experimenting with Kubernetes in 2016, when the platform was still in its infancy — “around version 1.6.5,” as Mashiach recalled.

“We were early adopters really focusing on building a platform inside Playtika,” he said. “It was very challenging, especially when things were not productized… everything was homegrown.”

Those early decisions, Rosenberg added, came with a price.

“When we talk about Playtika in the initial days… we needed three separate etcd clusters, one for Kubernetes events, one for Calico, another one for Kubernetes itself because you had nothing like Typha at that time. Everything, and I mean everything, was homegrown.”

By being so far ahead of the curve, Playtika accumulated deep technical debt. From custom Jenkins pipelines to forked Helm charts and IP exhaustion caused by routable pods, the team had to maintain a sprawling ecosystem that few outside engineers could even understand. “Nothing has been validated,” Rosenberg said bluntly. “It’s all YOLO.”

The wall every platform team hits

Eventually, the team realized that their internal engineering culture — the same one that had driven Playtika’s success — was also holding them back.

“We started to lock in within our own solution because we usually say, okay, we know it better,” said Mashiach. “If the vendor cannot produce it for us… let me build it myself. But as time goes by, we saw that as being much more a downside than an upside.”

Rosenberg put it in perspective:

“No one is that special. Every organization says, ‘We’re a unicorn, we’re unique, nothing will ever work for us.’ Small little notice — no one is that special.”

With thousands of bare metal servers, multiple data centers, and even a full depletion of IPv4 space, the Playtika team faced a scale problem that made every upgrade and patch cycle painful. The team tried Pivotal PKS and other solutions, but nothing worked at Playtika’s scale.

The turning point came when leadership backed a cultural shift. Mashiach described it as “management backup combined with engagement from the engineering teams.” That alignment allowed them to reimagine their platform with help from outside experts — and a strong foundation in open source.

The move to buy versus build

Rosenberg summarized the shift succinctly:

“The typical way it’s set is build versus buy. What we got pushed on from the management within Playtika is, no — we’re moving to buy versus build.”

That mindset doesn’t mean abandoning innovation; it means focusing on what differentiates you rather than reinventing what the industry has already solved.

“No company is a master of all trades,” Rosenberg said. “Buy if it doesn’t make sense, because it is too expensive or doesn’t solve my issues — then you build.”

TeraSky helped Playtika evaluate a wide range of options, ultimately settling on Spectro Cloud Palette as the backbone of their new platform.

“When we looked at different fleet manager solutions, what we ended up going on here with the best fit for Playtika was actually Spectro Cloud,” said Rosenberg. “It’s based off of Cluster API and brings in a bunch of open source projects — external secrets operator, external DNS, monitoring, observability — and it allowed us to build a full solution for our end users without getting locked into a specific vendor.”

Mashiach confirmed the results: “Actually working,” he said with a smile — a dry understatement that drew smiles from the audience.

Building a declarative future

The replatforming effort combined Spectro Cloud Palette with technologies like Crossplane, CloudStack, Kyverno, and FluxCD to create a fully declarative GitOps-driven environment.

“Everything is run via Git,” Mashiach explained. “There is no direct access to the clusters. Nobody is allowed to deploy anything directly… Version control, push to Flux, Flux pushes this downstream and everything is getting deployed.”

This new model gave Playtika both consistency and control — while freeing their engineers from the manual toil that had slowed them down. Time to provision clusters dropped from “a day, day and a half, to maybe an hour,” according to Mashiach. 

The power of partnership

A recurring theme throughout the talk was partnership — between Playtika and TeraSky, and between both companies and vendors like Spectro Cloud.

“Cooperation or collaboration with other people… gives you clarity,” said Mashiach. “I challenge everybody every day, constantly. But Scott is doing the same for me, which is excellent.”

That outside perspective was key to designing a solution grounded in industry best practices rather than legacy assumptions. “Everything was measured, everything was tested,” Mashiach emphasized. “We know what we are doing based on those environments.”

Why this story matters

Playtika’s journey mirrors what many enterprise platform teams are experiencing: the realization that scale and sophistication can become barriers to progress if they’re built on fragile foundations. The message from TeraSky and Playtika is clear — move from duct tape to declarative.

Rosenberg’s takeaway summed it up best:

“Early adoption does cause technical debt. That doesn’t mean we should not be early adopters. It means we have to plan how to get off of that later on… 

Maschiach jumps in with the last word: “Run fast, fail fast.”

Visit us at KubeCon

Whether you’re managing a handful of clusters or thousands, it’s time to move from duct tape to declarative. If you’re at KubeCon + CloudNativeCon North America, stop by the Spectro Cloud booth 621 this week to learn how our platform — the same one powering Playtika’s modern platform journey — can fit into your platform engineering stack, working with visionary partners like TeraSky.