3 ms·
> How costly can it be to test the file fully in a CI job? It didn't need a CI job. It just needed one person to actually boot and run a Windows instance with
by denton-scratch 2y ago
> How costly can it be to test the file fully in a CI job?
It didn't need a CI job. It just needed one person to actually boot and run a Windows instance with the Crowdstrike software installed: a smoke test.
TFA is mostly an irrelevent discourse on the product architecture, stuffed with proprietary Crowdstrike jargon, with about a couple of paragraphs dedicated to the actual problem; and they don't mention the non-existence of a smoke test.
To me, TFA is not a signal that Crowdstrike has a plan to remediate the problem, yet.
- hrpnk 2y agoThey mentioned they do dogfooding. Wonder why it did not work for this update.
- xh-dude 2y agoThey discuss dogfooding “Sensor Content”, which isn’t “Rapid Response Content”. Overall the way this is written up suggests some cultural problems.
- stefan_ 2y agoYou just got tricked by this dishonest article. The whole section that mentions dogfooding is only about actual updates to the kernel driver. This was not a kernel driver update, the entire section is irrelevant. This was a "content file", and the first time it was interpreted by the kernel driver was when it was pushed to customer production systems worldwide. There was no testing of any sort.
- broknbottle 2y agoAll these people claiming they didn’t have canaries. They actually did but people are in denial that they are the canary for crowdstrike lol
- cjbprime 2y agoIt's worse than that -- if your strategy actually was to use the customer fleet as QA and monitoring, then it probably wouldn't take you an hour and a half to notice that the fleet was exploding and withdraw the update, as it did here. There was simply no QA anywhere.