Bcachefs hangs on intense I/O. #787
Comments
|
Kernel messages (system hangs) More kernel messages Again kernel messages |
|
It hangs again. Running fsck few times doesnt help. The bug is clearly reproducible.
Sorry for the photos, but I need to write from mobile phone, as computer unresponsive now |
|
Need to add: after switching to kernel 6.12.0 the problem fixed. Also I found, that at the first boot file system have been migrated to new version. |
|
Too early. On the second try (building AOSP) it hangs again! Have no logs for now. Moreover, situation is worse than before. With new kernel I have no graphics driver (nvidia...), and unable to load previous kernel, as it becomes converting filesystem back, and latter is never ending process. With new (6.12.0) kernel it works strange now. No more effect of the caching from SSD. All is VERY SLOW and data almost always read from HDD, not from SSD. |
|
I have evacuated, removed, added again SSD drive. Now caching works (I see in dstat output). But now I have following error messages in dmesg: |
|
Another message, cleaner one: I just do commands |
|
Looks like re-adding SSD drive fixed the problem. At least, build is running now with no problems. |
|
I've been seeing the same thing on kernel 6.11.7 |
|
I can pretty routinely hang the FS if I thrash the disk hard enough (which is easy to do when backing files up to it via |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Basically, the issue is the same as #673, but I cannot reopen the latter issue.
I have attached dmesg to the issue. In my particular case, the rebuild of AOSP project caused these messages, at least.
Also, the system may become completely unresponsive with load average like 2000. New processes cannot be started in this case, all stuck at IO. I will try to add a few photos I made from the screen. What is interesting, in this state existing processes (like web browser) work fine.
The kernel version is 6.10.1. Previously, @koverstreet closed the bug #637 with the words, what "if anyone is still seeing this on 6.10 go ahead and reopen". Of course, anybody cannot reopen the bug at Github... But the bug still persists.
Need to add, that bug is definitely depending on filesystem state: the problem doesn't appear in the past.
Another problems: out of kernel fsck fixes the same errors every time (like absent or zero lru, I don't remember details and didn't take the screenshot). In-kernel fsck hangs too.:-(
Looks, I need to copy all the data to another disk. I want to ask suggestion, how should I do this: currently I have one HDD and one SSD for bcachefs. Both fully occupied by one file system, but SSD is used for cache, see
superblock.txt Basically, I want to create a new file system on another HDD, copy all the data to it, and when add first HDD as a mirror and add (single, I haven't two) SDD as a cache. How can I do this in correct way? PS: I understood, what I have problems if SDD fails, but at first I need performance, but not reliability.
dmesg.txt.gz
The text was updated successfully, but these errors were encountered: