HAMMER goes to nirvana
Antonio Huete Jiménez
tuxillo at quantumachine.net
Sat Sep 27 02:34:54 PDT 2014
Quoting Andrey Oktyabrskiy <ano at bestmx.net>:
> On 21.09.2014 19:23, Andrey Oktyabrskiy wrote:
>> On 21.09.2014 18:46, Justin Sherrill wrote:
>>> On Sun, Sep 21, 2014 at 8:20 AM, Andrey Oktyabrskiy <ano at bestmx.net>
>>>> My home server cannot mount one of the hammer filesystems. I see 4
>>>> messages and OS does not respond any more:
>>> The 'man 5 hammer' page mentions it's possible to turn off redo by
>>> setting vfs.hammer.skip_redo to 1 (mount despite errors) or 2 (don't
>>> even try redo).
>> Thank you, I'll try it when will make backup...
> There is no vfs.hammer.skip_redo sysctl.
It's not a sysctl but a tunable. I.e. something to put on /boot/loader.conf
From hammer(5) manpage:
In the unlikely case HAMMER mount fails due redo recovery (stage 2
recovery) being corrupted, a workaround to skip this stage can be applied
by setting the following tunable:
Possible values are:
0 Run redo recovery normally and fail to mount in the case of error
1 Run redo recovery but continue mounting if an error appears.
2 Completely bypass redo recovery.
Related commands: mount_hammer(8)
>>> If that doesn't get anywhere, you can boot to single user mode (one of
>>> the boot menu options) or boot from a LiveCD/USB image, and try
>>> mounting the volume then. If you can get it online, 'hammer show -f
>>> /dev/thatdrivedevice | grep B' will I think filter for errors. See
>>> man 8 hammer. I have not tried this directly, so your mileage may
>> Yes, I can't get it online in single user mode too.
>> I switched to 3.8.2, recompiled world and kernel and can mount that FS!
>> So this bug is in the recent -master (compiled yesterday).
> 3.8.2 stopped to work with the same symptoms. I installed 3.6 and it
> works fine with that hammer filesystem. I successfully back up my
> data and ready for some experiments with that FS.
> At the moment there is 3.9 from yesterday master. I formatted that
> partition and tried to write-delete-write large amount of data
> there. And have exactly the same behaviour - system hangs. What
> information can I provide for debugging?
More information about the Users