Re: ceph and ext4

[Martin Mailand <martin@tuxadero.com>]

so snapshots of rbd images would be safe? or would they hit the limit?

Sage Weil schrieb:
> On Mon, 14 Nov 2011, Gregory Farnum wrote:
>> On Mon, Nov 14, 2011 at 5:33 AM, Martin Mailand <martin@tuxadero.com> wrote:
>>> Hi Tomasz,
>>> as far as I know it still has this limit.
>>> But it should be relatively safe to use it.
>>>
>>> http://marc.info/?l=ceph-devel&m=131942130322957&w=2
>>>
>>> If we hit the 4KB limit of xattrs in ext4 how does it show up in the rbd
>>> layer?
>>>
>>> How does it show up in the fs layer, would the fs still be clean?
>> Right now it would show up very badly, unfortunately. (And yes, the
>> limit is still there.) You'd notice, though you might manage to
>> corrupt some of your data first. :/
> 
> Well, the osd's are now more careful about being fail-stop, so if they hit 
> the xattr limit they crash.  So there won't be data corruption per se, 
> except that you won't be able to start the OSD up again because the 
> journal replay will keep hitting the limit.
> 
>> However, if you're not taking snapshots and you're not using xattrs
>> yourself, you won't hit it with rbd or the Ceph FS.
> 
> Right.  Nothing sets large xattrs on objects in rbd.  For the file system, 
> this would only happen on extremely (!) deeply nested directories (ceph 
> dfs xattrs are managed by the MDS, not as object attrs).
> 
> sage
> 
> 
>> -Greg
>>
>>> -martin
>>>
>>>
>>> Am 14.11.2011 14:09, schrieb Tomasz Paszkowski:
>>>> what about limit on xattr size ? Is it still limited to 4KB ?
>>>>
>>>>
>>>>
>>>> On Mon, Nov 14, 2011 at 1:15 PM, Martin Mailand<martin@tuxadero.com>
>>>>  wrote:
>>>>> Hi Christian,
>>>>> I am not sure if you noticed, but your ext4 bug is fixed in mainline. I
>>>>> am
>>>>> running a ceph cluster with 40+ vms for over a week by now, without any
>>>>> problems. An fsck.ext4 shows the ext4 is clean.
>>>>> The performance of ext4 is much better than btrfs, no rise in the load of
>>>>> the osd's.
>>>>>
>>>>> -martin
>>>>> --
>>>>> To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
>>>>> the body of a message to majordomo@vger.kernel.org
>>>>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>>>>>
>>>>
>>>>
>>> --
>>> To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
>>> the body of a message to majordomo@vger.kernel.org
>>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>>