Clever Geek Handbook

Copying volumes to storage systems via Linux server using XCOPY

It happens that you need to get a complete copy of a volume within one storage system (DSS), not a snapshot, a clone, but a full-fledged volume. But the storage system does not always allow you to do this within yourself by your own means. It seems like the only option is to copy through the server, but at the same time the entire amount of data will be chased through the server itself, the network to the storage system and storage ports, loading all these components. But there are SCSI commands that can allow you to do everything within the framework of the storage system itself, and if your system supports VAAI from VMware, then almost 100%, which is supported by the XCOPY (EXTENDED COPY) command, which tells the array what to copy and where, without involving this process server and network.



It seems like everything should be simple, but I didn't find ready-made scripts right away, I had to reinvent the wheel. Linux was chosen for the server OS, and the ddpt command (http://sg.danny.cz/sg/ddpt.html) was chosen as the copy tool. Any volumes from any OS can be copied using this combination, since copying is block-by-block on the storage side. Since it is necessary to copy block-by-block, and the number of blocks must be counted, the blockdev command was used to count the number of such iterations. The maximum block size was obtained empirically, with a large block ddpt did not work in fact. The result is the following rather simple script: 



#!/bin/bash
# first parameter = input device
# second parameter = output device
# device size must be the same
# changing bs variable can reduce speed, max speed should be at bs=32768. 32768 is max setting, lower settings should be calculated dividing by 2

set -o nounset
bs=32768
s=`blockdev --getsz $1`
i=0
while [ $i -le $s ]
do
ddpt of=$2 bs=512 oflag=xcopy,direct if=$1 iflag=xcopy,direct count=$bs verbose=-1 skip=$i seek=$i
i=$(( $i+$bs ))
done


Let's do a little check! Well, how small, 1TB file was created and checked by md5sum not quickly :) 



root@sales-demo-05:/home/vasilyk# blockdev --getsz /dev/mapper/mpathfs
2516582400
root@sales-demo-05:/home/vasilyk# blockdev --getsz /dev/mapper/mpathfr
2516582400
root@sales-demo-05:/home/vasilyk# mount /dev/mapper/mpathfs /xcopy_source/
mount: /xcopy_source: wrong fs type, bad option, bad superblock on /dev/mapper/mpathfs, missing codepage or helper program, or other error.
root@sales-demo-05:/home/vasilyk# mkfs /dev/mapper/mpathfs
mke2fs 1.44.1 (24-Mar-2018)
Discarding device blocks: done
Creating filesystem with 314572800 4k blocks and 78643200 inodes
Filesystem UUID: bed3ea00-c181-4b4e-b52e-d9bb498be756
Superblock backups stored on blocks:
        32768, 98304, 163840, 229376, 294912, 819200, 884736, 1605632, 2654208,
        4096000, 7962624, 11239424, 20480000, 23887872, 71663616, 78675968,
        102400000, 214990848

Allocating group tables: done
Writing inode tables: done
Writing superblocks and filesystem accounting information: done

root@sales-demo-05:/home/vasilyk# mount /dev/mapper/mpathfs /xcopy_source/
root@sales-demo-05:/home/vasilyk# ls -l /xcopy_source/
total 16
drwx------ 2 root root 16384 Aug 19 15:35 lost+found
root@sales-demo-05:/home/vasilyk# head -c 1T </dev/urandom > /xcopy_source/1TB_file
root@sales-demo-05:/home/vasilyk# ls -l /xcopy_source/
total 1074791444
-rw-r--r-- 1 root root 1099511627776 Aug 19 17:25 1TB_file
drwx------ 2 root root         16384 Aug 19 15:35 lost+found
root@sales-demo-05:/home/vasilyk# umount /xcopy_source
root@sales-demo-05:/home/vasilyk# mount /dev/mapper/mpathfr /xcopy_dest/
mount: /xcopy_dest: wrong fs type, bad option, bad superblock on /dev/mapper/mpathfr, missing codepage or helper program, or other error.
root@sales-demo-05:/home/vasilyk# cat xcopy.sh
#!/bin/bash
# first parameter = input device
# second parameter = output device
# device size must be the same
# changing bs variable can reduce speed, max speed should be at bs=32768. 32768 is max setting, lower settings should be calculated dividing by 2

bs=32768
s=`blockdev --getsz $1`
i=0
while [ $i -le $s ]
do
ddpt of=$2 bs=512 oflag=xcopy,direct if=$1 iflag=xcopy,direct count=$bs verbose=-1 skip=$i seek=$i
i=$(( $i+$bs ))
done
root@sales-demo-05:/home/vasilyk# time ./xcopy.sh /dev/mapper/mpathfs /dev/mapper/mpathfr
real    11m30.878s
user    2m3.000s
sys     1m11.657s


What was happening on the storage system at this moment: Let's



image

continue with Linux. 



root@sales-demo-05:/home/vasilyk# mount /dev/mapper/mpathfr /xcopy_dest/
root@sales-demo-05:/home/vasilyk# ls -l /xcopy_dest/
total 1074791444
-rw-r--r-- 1 root root 1099511627776 Aug 19 17:25 1TB_file
drwx------ 2 root root         16384 Aug 19 15:35 lost+found
root@sales-demo-05:/home/vasilyk# mount /dev/mapper/mpathfs /xcopy_source/
root@sales-demo-05:/home/vasilyk# md5sum /xcopy_source/1TB_file
53dc6dfdfc89f099c0d5177c652b5764  /xcopy_source/1TB_file
root@sales-demo-05:/home/vasilyk# md5sum /xcopy_dest/1TB_file
53dc6dfdfc89f099c0d5177c652b5764  /xcopy_dest/1TB_file
root@sales-demo-05:/home/vasilyk#




Everything worked out, but test and use at your own peril and risk! It is better to take snapshots as a source volume, for a start.


More articles:


All Articles