Closed MateusRodCosta closed 1 year ago
I could blame on rpm-ostree
Sorry, it's going to take significant evidence that somehow the I/O patterns created by rpm-(ostree) trigger this.
Please try other tools for testing filesystem I/O - for example fio.
If you do gather more information and it really points towards (rpm-)ostree, it's OK to reopen.
Sorry, it's going to take significant evidence that somehow the I/O patterns created by rpm-(ostree) trigger this.
Hi, I will investigate further in the future, for now I just disabled Automatic Updates via Gnome Software.
I believe that in at least two of the intances of the issue I was also programming at the time the corruption occurred. So, that would be rpm-ostree + VS Code + Flutter + Android AVD all at once.
Anyway, I will re-open if I find more info.
@cgwalters Any chance https://github.com/ostreedev/ostree/pull/2874 was related to my issue?
The instances where file corruption always seemed to happen on one of rpm-ostree update attempts, maybe it was ostree itself instead of rpm-ostree IO usage?
It seems ostree 2023.4 fixes it, so I guess if I update and re-enable automatic system updates I should be fine.
Of note, before I switched SSDs and formatted the laptop, the files which got corrupted due to rpm-ostree were some files related to the ostree commits (from isnide /ostree
). The one file that got corrupted after the SSD replacement (the one from the logs in this issue) was some random game image stored in my home folder, luckly I could easily replace it.
On Fri, Jun 23, 2023, at 6:44 PM, Mateus Rodrigues Costa wrote:
@cgwalters https://github.com/cgwalters Any chance ostreedev/ostree#2874 https://github.com/ostreedev/ostree/pull/2874 was related to my issue?
No; that could cause different files to appear, but not file system corruption.
Basically file system corruption can only happen through file system (or other kernel) bugs or faulty hardware. Those bugs could be triggered by userspace bugs, but not caused by.
Host system details
Provide the output of
rpm-ostree status
.Expected vs actual behavior
Apparently rpm-ostree auto update causes btrfs corruption sometimes, this is usually noticed because the filesystem turns read only close to the time that I notice the fan of the laptop starts running very fast. This usually happens because GNOME Software decided to trigger auto update after I resume the laptop from suspend.
Logs:
Expected:
No broken filesystem.
Steps to reproduce it
Let Gnome Software update your system.
Would you like to work on the issue?
No, I don't have the necessary knowledge.
Extra notes
I use an NVME SSD (previously an ADATA that came with the laptop, now a WD Black that I bought due to thinking it was a defective SSD). IIRC, I believe I hit this issue 4 times, where only the last 3 ones I could blame on rpm-ostree, and two of those were ostree related files that got corrupted.
I don't know, but maybe rpm-ostree is writing to the disk a t full speed and, due to the NVME SSD not having a limit, it might be part of what causes it. (It could be the SSD running hot or the btrfs not being to handle the speed?)