[dm-devel] Does dm-zoned support buffered write?

[Ming Lin]

On Mon, May 15, 2023 at 4:46 AM Damien Le Moal <dlemoal at kernel.org> wrote:
> >>
> >> ===
> >> At another terminal,
> >>
> >> root at smr_dev:~# ps aux | grep mkfs.ext4
> >> root     1411791  2.8  0.0  30992 19864 pts/1    D+   01:30   0:01
> >> mkfs.ext4 /dev/dm-0
> >> root     1413640  0.0  0.0  13972  2496 pts/0    S+   01:31   0:00
> >> grep mkfs.ext4
> >>
> >> root at smr_dev:~# cat /proc/1411791/stack
> >> [<0>] wait_on_page_bit+0x133/0x270
> >> [<0>] wait_on_page_writeback+0x25/0x70
> >> [<0>] __filemap_fdatawait_range+0x86/0xf0
> >> [<0>] file_write_and_wait_range+0x74/0xb0
> >> [<0>] blkdev_fsync+0x16/0x40
> >> [<0>] do_fsync+0x38/0x60
> >> [<0>] __x64_sys_fsync+0x10/0x20
> >> [<0>] do_syscall_64+0x2d/0x70
> >> [<0>] entry_SYSCALL_64_after_hwframe+0x44/0xa9
>
> Not sure if this is a bug, but doing a simple mkfs.ext4 on dm-zoned with a large
> SMR disk can take *a very loooooong* time. This is because mkfs.ext4 does a lot
> of random writes all over the place. So just running that, dm-zoned goes into
> heavy GC mode...
>
> To speed things up (and improve runtime performance), use the packed-metadata
> format: mkfs.ext4 -E packed_meta_blocks=1
> Or do a mkfs.xfs to compare and see how much faster it is.

My SMR disk is 24T, 969 conventional zones + 95891 sequential zones,
each zone is 256M.

Amazingly, mkfs.xfs only takes ~8 seconds and mkfs.ext4 -E
packed_meta_blocks=1 only takes 32 seconds.
Thanks!

root at smr_dev:~# time mkfs.xfs /dev/dm-3
meta-data=/dev/dm-3              isize=512    agcount=24, agsize=268435455 blks
         =                       sectsz=4096  attr=2, projid32bit=1
         =                       crc=1        finobt=1, sparse=1, rmapbt=0
         =                       reflink=0
data     =                       bsize=4096   blocks=6346375168, imaxpct=5
         =                       sunit=0      swidth=0 blks
naming   =version 2              bsize=4096   ascii-ci=0, ftype=1
log      =internal log           bsize=4096   blocks=521728, version=2
         =                       sectsz=4096  sunit=1 blks, lazy-count=1
realtime =none                   extsz=4096   blocks=0, rtextents=0

real    0m8.834s
user    0m0.002s
sys     0m0.322s


root at smr_dev:~# time mkfs.ext4 -F -E packed_meta_blocks=1 /dev/dm-3
mke2fs 1.44.5 (15-Dec-2018)
/dev/dm-3 contains a xfs file system
Discarding device blocks: done
Creating filesystem with 6346375168 4k blocks and 396648448 inodes
Filesystem UUID: 93136647-0214-4ea2-8403-2561a4cd3e42
Superblock backups stored on blocks:
        32768, 98304, 163840, 229376, 294912, 819200, 884736, 1605632, 2654208,
        4096000, 7962624, 11239424, 20480000, 23887872, 71663616, 78675968,
        102400000, 214990848, 512000000, 550731776, 644972544, 1934917632,
        2560000000, 3855122432, 5804752896

Allocating group tables: done
Writing inode tables: done
Creating journal (262144 blocks): done
Writing superblocks and filesystem accounting information: done


real    0m32.449s
user    0m0.320s
sys     0m1.165s