CrowdStrike recently caused a widespread Blue Screen of Death (BSOD) issue on Windows PCs, disrupting various sectors. However, this was not an isolated incident, CrowdStrike affected Linux PCs also.
As an application developer (rather than someone who can/does code operating systems) I was just left open-mouthed …
Looks like they’re delivering “code as content” to get around the rigour of getting an updated driver authorised by MS. I realise they can’t wait too long for driver approval for antivirus releases but surely - surely - you have an ironclad QA process if you’re playing with fire like this.
Setting the update policy to N-2 (or any other configuration) would not have avoided the issue. The Falcon sensor itself wasn’t updated, which is what the update policy controls. As it turns out, you cannot control the content channel updates - you simply always get the updates.
I read on another thread that an admin was emulating a testing environment by blocking CrowdStrike IPs on their firewall for the whole network before each update, with the exception of a couple machines. It’s stupid that he has to do this but hey, his network was unaffected
We would if we were able to control their “deployable content”.
Serious question, can you not? There isn’t an option to…like…set a review system first?
For antivirus definitions? No, and you wouldn’t want to.
But it sounds like this added files / drivers or something, not just antivirus rules?
Turns out it was a content update that caused the driver to crash but the update itself wasn’t a driver (as per their latest update.)
Oh, wow.
Found this post that explains what happened in detail: https://lemmy.ohaa.xyz/post/3522666
As an application developer (rather than someone who can/does code operating systems) I was just left open-mouthed …
Looks like they’re delivering “code as content” to get around the rigour of getting an updated driver authorised by MS. I realise they can’t wait too long for driver approval for antivirus releases but surely - surely - you have an ironclad QA process if you’re playing with fire like this.
Do you know if the sensor update policy had been set to N-2 would this have avoided the issue?
No it would not.
Setting the update policy to N-2 (or any other configuration) would not have avoided the issue. The Falcon sensor itself wasn’t updated, which is what the update policy controls. As it turns out, you cannot control the content channel updates - you simply always get the updates.
💀 Fucking hell CrowdStrike.
https://nitter.poast.org/patrickwardle/status/1814343502886477857
I read on another thread that an admin was emulating a testing environment by blocking CrowdStrike IPs on their firewall for the whole network before each update, with the exception of a couple machines. It’s stupid that he has to do this but hey, his network was unaffected
Minimum safe distance.