New issue
Advanced search Search tips

Issue 866751 link

Starred by 6 users

Issue metadata

Status: Fixed
Owner:
Closed: Dec 13
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Chrome
Pri: 1
Type: Bug



Sign in to add a comment

Unable to start termina after a crash

Project Member Reported by tbuck...@chromium.org, Jul 24

Issue description

After Chrome crash, "vmc start termina" no longer works.

Feedback report: https://listnr.corp.google.com/report/85542701135
 
Cc: smbar...@chromium.org
Seems like the btrfs disk was corrupted:

2018-07-09T19:55:03.828690-07:00 INFO VMBOOT(3)[12367]: [    0.932753] BTRFS error (device vdb): parent transid verify failed on 632291328 wanted 13702 found 13575#015
2018-07-09T19:55:03.830384-07:00 INFO VMBOOT(3)[12367]: [    0.934491] BTRFS error (device vdb): parent transid verify failed on 632291328 wanted 13702 found 13575#015
2018-07-09T19:55:03.831136-07:00 INFO VMBOOT(3)[12367]: [    0.935727] BTRFS warning (device vdb): failed to read tree root#015
2018-07-09T19:55:03.849580-07:00 INFO VMBOOT(3)[12367]: [    0.954235] BTRFS error (device vdb): open_ctree failed#015
2018-07-09T19:55:03.857473-07:00 ERR vm_concierge[12321]: Stateful setup failed
2018-07-09T19:55:03.860089-07:00 ERR VM(3)[12322]:  stateful_setup: Failed to mount stateful disk


2018-07-10T10:34:56.698403-07:00 ERR VM(6)[12322]: kernel: BTRFS error (device vdb): parent transid verify failed on 632291328 wanted 13702 found 13575
2018-07-10T10:34:56.698408-07:00 ERR VM(6)[12322]: kernel: BTRFS error (device vdb): parent transid verify failed on 632291328 wanted 13702 found 13575
2018-07-10T10:34:56.698414-07:00 WARNING VM(6)[12322]: kernel: BTRFS warning (device vdb): failed to read tree root
2018-07-10T10:34:56.698419-07:00 ERR VM(6)[12322]: kernel: BTRFS error (device vdb): open_ctree failed


Part of the problem is  issue 835035  where the guest kernel panics if chrome restarts, which means there isn't enough time for the file system to properly flush all of its dirty pages.   I think we might also be too aggressive in killing off device processes when the VM exits.  In theory, since btrfs is a copy-on-write file system it should handle system crashes just fine.  However, if the block device process gets killed before it's had a chance to finish writing out all of its blocks then it makes sense that the underlying fs ends up in a corrupted state.
we should keep an eye out for more reports like this. This one might be from before we fixed the guest flush command.

Comment 3 Deleted

I also have a similar problem. I was installing LibreOffice when it crashed and I am unable to start it again. I'm not frustrated because the container is not starting, but because of the fact that I am unable to retrieve my files. 
Labels: -M-69
<triage>@chirantan should we close this? Removing milestone since M69 has past.</triage>
Status: Fixed (was: Assigned)
This was likely fixed when we landed the sync changes for the block device.

Sign in to add a comment