Unable to recover hammer volume.

jscottkasten at yahoo.com jscottkasten at yahoo.com
Mon Jan 16 07:04:21 PST 2017

----- Original Message -----

From: Tomohiro Kusumi <kusumi.tomohiro at gmail.com>

>I don't think you can attempt to remove those in your situation.
>But assuming your corrupted sectors as proper inodes and go on to rm
>those files isn't really safe operation either. Since your ondisk
>inodes are known to be corrupted from those dmesg lines, it's
>possible/likely that something unexpected occurs during unlinking and
>eventually hits assertion or destroy irrelevant sectors.

I had considered that too.  A broken chain can leave isolate pieces everywhere.

>I think what's really missing in hammer1 (in terms of corruption) is
>fsdb type program and fsck. Some (or many) filesystems do have fsck
>even if they have some sort of journal equivalent.

fsck and friends do some significant scanning.  While it is sometimes possible to repair things, I've also seen more than my share of things that are just beyond such tools.

So I should probably just be happy that the file system knows exactly what's broken and what isn't.  File systems without these CRC checks will happily feed you garbage.


2017-01-16 23:27 GMT+09:00  <jscottkasten at yahoo.com>:
> ----- Original Message -----
> From: Tomohiro Kusumi <kusumi.tomohiro at gmail.com>
>>hammer_btree_extract: CRC DATA @ 9000000a4a1f3280/128 FAILED
> Yes, I have this, and "critical error" reading inodes.
>>This means bunch of inodes in certain sectors nearby were somehow
>>corrupted. These are often found only when those inodes are read the
>>next time after corruption, thus doesn't necessary affect booting
>>unless those inodes are read on booting.
> Yes, this is a backup volume.
>>Onboot recovery really has nothing to do with random broken sectors
>>where undo fifo is not involved, nor is it a problem of filesystems in
>>general including hammer1. Also note that "recovery" here doesn't mean
>>hammer1 somehow magically fixes user data. It means the filesystem
>>keeps consistency of itself.
> I would be happy to just delete them somehow.  I can rsync any incomplete data easily enough.
>>Having said that, hammer1 could give an option for users to at least
>>do whatever they want to do on a broken fs. It currently forces
>>readonly mount, kernel panic, etc.
> And there in lies the problem.  The forced read only mount prevents me from doing anything to fix it...  :-(
> I just did check the mount_hammer page to see if there were any useful options, but nothing unfortunately.
> Thanks,
> -S-
> 2017-01-16 22:52 GMT+09:00 PeerCorps Trust Fund <ipc at peercorpstrust.org>:
>> Hi,
>> It sounds pretty much like its supposed to operate if your file system tree
>> is potentially corrupted somehow. You may very well need to copy off the
>> data and recreate the file system. Unless of course the underlying block
>> storage has some issues (bad sectors etc.).
>> I do recall there not yet being a way to gracefully fix a "broken" HAMMER
>> file system.
>> Perhaps I am wrong and someone else will offer more enlightened feedback.
>> Out of curiosity when the file system was mounted at boot, did it not fail
>> to mount? It seems like it should have been brought back to a consistent
>> state after your reboot.
>> Mike
>> On 01/16/2017 03:32 PM, jscottkasten at yahoo.com wrote:
>>> Hi guys,
>>> I've tried searching the answer for this, but cannot find any proper
>>> discussion on the topic.
>>> I have a hammer volume that had to be forced into an unclean shutdown.
>>> There are some files and/or folders that now have CRC errors.  When the
>>> damaged data is encountered, it causes hammer to report a critical error and
>>> remount the file system as READ ONLY.  Thus, I am unable to delete or remove
>>> the bad data in any way to recover the file system to an operable state.
>>> Is there some tool that I should run on the file system to do this???  I
>>> know the bad data is restricted to a small number of folders.  Would love to
>>> somehow purge it so the volume can be returned to normal operation.  It's a
>>> 2TB volume, so I'm hoping I don't have to reformat and re-create the volume.
>>> Thanks for any advice.
>>> Regards,
>>> -J. Scott Kasten-

More information about the Users mailing list