Repository navigation
Permission issues on Windows host #4020
Description
https://ci.nodejs.org/computer/test%2Dazure%5Fmsft%2Dwin11%5Fvs2022%2Dx64%2D4/builds
00:00:00.467 > git clean -dffx
00:00:00.467 fatal: detected dubious ownership in repository at 'C:/workspace/node-test-binary-windows-js-suites/build'
00:00:00.467 'C:/workspace/node-test-binary-windows-js-suites/build' is owned by:
00:00:00.467 BUILTIN/Administrators (S-1-5-32-544)
00:00:00.467 but the current user is:
00:00:00.467 node-11-22-4/bdirhlxvsgdl (S-1-5-21-1229986324-1897981146-1831166073-500)
I put the agent offline for now.
Activity
StefanStojanovic commentedon Feb 13, 2025
Hmm this is the same machine as here, It was fixed, but I assume it rebooted or something like that and this is back. I'll take another look.
StefanStojanovic commentedon Feb 13, 2025
So as it turns out, this all started after running some tests from a
nodejs-private/node-privatePR 664. I do not have access to that repo, but based on what we're facing here it somehow stripped the user of his privileges. I'll look into this more. leaving the node offline during that time.targos commentedon Feb 13, 2025
That PR was the v20.18.2 release proposal
StefanStojanovic commentedon Feb 13, 2025
Upon comparing it to the other Win11 VS2022 machines I noticed there were a few additional security policies enabled on that machine. I removed them and rebooted and all seems to be OK now. I also manually cleaned the workspace directory because of the ownership issues. Let's keep this issue open for some more time, but it should be resolved now.
richardlau commentedon Feb 15, 2025
I think this is back (#4022) 😞. I've taken the machine back offline from Jenkins.
StefanStojanovic commentedon Feb 17, 2025
Not really back, as this is another machine, but the same thing, I'll go through all the 3 other Win11 x64 machines to check it out
StefanStojanovic commentedon Feb 17, 2025
So here is the status. Of 4 machines 2 are affected (1 was last week but I resolved it). Those are machines 2 and 3. I fixed those and reenabled them. They should work now. Machine 1 is still not affected by this and I expect it to get affected at some point because this is probably due to some update as no manual actions were taken to enable this on machines and last week machines 2 and 3 didn't have that problem.
P.S. I'm OOF this week on vacation. Luckily, I'm leaving tonight so I was able to fix this. If the same happens on machine 1 someone should just disable it and the 3 other machines should be enough to handle running tests. I'll check this in the afternoon to see if it is resolved, but based on last week's experience it should be.
StefanStojanovic commentedon Feb 24, 2025
Closing this since it hasn't shown up since my last comment. When this hits the last machine I'll apply the patch to it as well.