[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-bcache
Subject:    Re: [RFC] Live resize of backing device
From:       mingzhe <mingzhe.zou () easystack ! cn>
Date:       2023-02-28 2:46:18
Message-ID: ee891d7d-df93-e170-4ece-37d115c0fab2 () easystack ! cn
[Download RAW message or body]



在 2023/2/28 06:08, Eric Wheeler 写道:
> On Mon, 20 Feb 2023, mingzhe wrote:
>> 在 2023/2/19 17:39, Coly Li 写道:
>>>> Subject: [PATCH v2] bcache: Add support for live resize of backing devices
>>>>
>>>> Signed-off-by: Andrea Tomassetti <andrea.tomassetti-opensource@devo.com>
>>>
>>> Hi Andrea,
>>>
>>> I am fine with this patch and added it in my test queue now. Do you have an
>>> updated version, (e.g. more coding refine or adding commit log), then I can
>>> update my local version.
>>>
>>>
>> Hi, Coly
>>
>> I posted some patchsets about online resize.
>>
>> -[PATCH v5 1/3] bcache: add dirty_data in struct bcache_device
>> -[PATCH v5 2/3] bcache: allocate stripe memory when partial_stripes_expensive
>> is true
>> -[PATCH v5 3/3] bcache: support online resizing of cached_dev
>>
>> There are some differences:
>> 1. Create /sys/block/bcache0/bcache/size in sysfs to trigger resize
> 
> Can the final version name the sysfs entry "resize", because "size" sounds
> like you are setting a specific size, not triggering a resize.
> 
> --
> Eric Wheeler
> 
OK. I will update a version.

mingzhe
> 
> 
>> 2. Allocate stripe memory only if partial_stripes_expensive is true
>> 3. Simplify bcache_dev_sectors_dirty()
>>
>> Since the bcache superblock uses some sectors, the actual space of the bcache
>> device is smaller than the backing. In order to provide a bcache device with a
>> user-specified size, we need to create a backing device with a larger space,
>> and then resize bcache. So resize can specify the size is very necessary.
>>
>>
>>
>>
>>>
>>>
>>>> ---
>>>> Hi Coly,
>>>> this is the second version of the patch. As you correctly pointed out,
>>>> I implemented roll-back functionalities in case of error.
>>>> I'm testing this funcionality using QEMU/KVM vm via libvirt.
>>>> Here the steps:
>>>>    1. make-bcache --writeback -B /dev/vdb -C /dev/vdc
>>>>    2. mkfs.xfs /dev/bcache0
>>>>    3. mount /dev/bcache0 /mnt
>>>>    3. dd if=/dev/random of=/mnt/random0 bs=1M count=1000
>>>>    4. md5sum /mnt/random0 | tee /mnt/random0.md5
>>>>    5. [HOST] virsh blockresize <vm-name> --path <disk-path> --size
>>>>    <new-size>
>>>>    6. xfs_growfs /dev/bcache0
>>>>    6. Repeat steps 3 and 4 with a different file name (e.g. random1.md5)
>>>>    7. umount/reboot/remount and check that the md5 hashes are correct with
>>>>          md5sum -c /mnt/random?.md5
>>>>
>>>> drivers/md/bcache/super.c | 84 ++++++++++++++++++++++++++++++++++++++-
>>>> 1 file changed, 83 insertions(+), 1 deletion(-)
>>>>
>>>> diff --git a/drivers/md/bcache/super.c b/drivers/md/bcache/super.c
>>>> index ba3909bb6bea..1435a3f605f8 100644
>>>> --- a/drivers/md/bcache/super.c
>>>> +++ b/drivers/md/bcache/super.c
>>>>>
>>>
>>> [snipped]
>>>
>>>
>>
[prev in list] [next in list] [prev in thread] [next in thread]