We reduced a container image from 800GB to 2GB

Sealos says they slimmed a grotesquely bloated container—from a jaw-dropping 800GB down to 2GB—after mystery logs ballooned like a digital hoarder’s basement. The post explains that a log of failed logins swelled to 11GB and got copied across layers, multiplying the mess. They used a custom tool, image-manip, to “squash” layers (think: flattening stacks of changes) and surgically delete junk, and the cluster breathed again. Cue the comments: the loudest camp called it a “misconfig confession,” with one reader dismissing the write-up as basically “we broke it, we fixed it.” Another chorus piled on with “if your image is 800GB you’re doing something wrong,” turning the thread into a meme factory about “container weight loss plans” and “390:1 diets.” But defenders pointed out nuance: untrimmed explained the dev boxes need an SSH server because popular IDEs connect over SSH, which likely attracted bot storms that spammed failed logins and blew up logs. Then came the privacy panic: the idea of engineers peeking into customer images set off alarm bells about secrets and repos. Finally, the nerd police arrived: “read the article,” they scolded, noting the 11GB log got amplified by copy behavior—not a single 800GB image. Drama, receipts, and a very hungry log file. Read the full story at Sealos.