This post is more than 5 years old
13 Posts
0
1226
August 4th, 2009 14:00
Copying between file systems
Our Oracle group has had some problems when copying data from a snap shot to a file system that uses the host resources and causes performance issues. I would like to be able to copy between file systems within the Celerra rather than export and mount them to a host. After the copy is done, we would just mount that newly refreshed file system to the host and keep resource utilization within the Celerra when copying data. What is the path to the mounted file systems within the Celerra and is copying like this an option?
Thanks
Thanks
No Events found!


Rainer_EMC
6 Operator
•
8.6K Posts
0
August 5th, 2009 02:00
Assuming you have DART 5.6 and a Replicator license just select
Replications
New
Copy a File System
then in the "Choose a Checkpoint" box select which snapshot you want to be copied.
If you have a read-only file system of the same size you can specify that as the destination - otherwise the System will create one for you and names it _replica1
Note that the copy process creates a checkpoint on the destination fs - before you can delete it you need to delete the ckpt first
see attached screenshots
# ls -l FS_RAINER
total 8
drwxr-xr-x 2 root root 8192 Aug 5 10:40 lost+found
file system is empty
# ls -l FS_RAINER/.ckpt
total 8
drwxr-xr-x 5 root root 1024 Aug 5 10:42 2009_08_05_10.43.17_Europe_Berlin
# ls -l FS_RAINER/.ckpt/2009_08_05_10.43.17_Europe_Berlin/
total 16
drwxr-xr-x 2 root root 8192 Aug 5 10:40 lost+found
-rw-r--r-- 1 root root 56 Aug 5 10:42 RESTORE_ME.txt
but I've got a file to restore in a checkpoint
after the nas_copy from the GUI
# server_mount server_2 | grep RAI
FS_RAINER on /FS_RAINER uxfs,perm,rw
FS_RAINER_replica1 on /FS_RAINER_replica1 uxfs,perm,ro
I've got a new file system
# ls -l FS_RAINER_replica1/
total 16
drwxr-xr-x 2 root root 8192 Aug 5 10:40 lost+found
-rw-r--r-- 1 root root 56 Aug 5 10:42 RESTORE_ME.txt
which is a copy of the snapshot and can be exported
3 Attachments
copy_fs_3.jpg
copy_fs_1.jpg
copy_fs_2.jpg
dynamox
11 Legend
•
20.4K Posts
•
87.4K Points
0
August 4th, 2009 15:00
Eclipsys1
13 Posts
0
August 4th, 2009 15:00
Thanks
dynamox
11 Legend
•
20.4K Posts
•
87.4K Points
0
August 4th, 2009 16:00
here is an example of mine:
ckpt_fs107_Checkpoint_Daily_014 on /root_vdm_3/ckpt_fs107_Checkpoint_Daily_014 ckpt,perm,ro,accesspolicy=NATIVE
Eclipsys1
13 Posts
0
August 4th, 2009 16:00
Eclipsys1
13 Posts
0
August 4th, 2009 16:00
dynamox
11 Legend
•
20.4K Posts
•
87.4K Points
0
August 4th, 2009 17:00
Rainer_EMC
6 Operator
•
8.6K Posts
0
August 5th, 2009 01:00
on the Celerra itself (data mover - not control station) you can copy a complete file system or a snapshot to a new file system using nas_copy/fs_copy or the GUI
Please do NOT use the control station for the copying with Linux commands - its not a strong NFS client and it is crucial for the system operation.
If you want a single file copied you can try server_archive
I havent used it on a checkpoint but guess it would work
regards
Rainer
Rainer_EMC
6 Operator
•
8.6K Posts
0
August 5th, 2009 02:00
create an new file system of the same size:
# nas_fs -name FS_RAINER_COPY -create samesize=FS_RAINER pool=clar_r5_performance
mount it readonly
# server_mount server_2 -o ro FS_RAINER_COPY /FS_RAINER_COPY
server_2 : done
# server_mount server_2 | grep RAI
FS_RAINER on /FS_RAINER uxfs,perm,rw
FS_RAINER_COPY on /FS_RAINER_COPY uxfs,perm,ro
see what checkpoints you've got on the source file system:
# fs_ckpt FS_RAINER -l
id ckpt_name creation_time inuse fullmark total_savvol_used ckpt_usage_on_savvol
217 ckpt1 08/05/2009-10:43:18-CEST y 90% 9% 8%
then use nas_copy to copy that checkpoint to the destination file system:
# nas_copy -name test -source -ckpt ckpt1 -destination -fs FS_RAINER_COPY -interconnect loopback
OK
now its copied:
# ls -l FS_RAINER_COPY/
total 16
drwxr-xr-x 2 root root 8192 Aug 5 10:40 lost+found
-rw-r--r-- 1 root root 56 Aug 5 10:42 RESTORE_ME.txt
and you can just export it and use it
Rainer_EMC
6 Operator
•
8.6K Posts
0
August 5th, 2009 03:00
Its usage syntax is a bit awkward IMHO but it has the advantage that like NDMP or nas_copy it runs on
the data mover and doesnt use the control station CPU or CS-DM internal network
for example:
figure out where my checkpoint is mounted:
# server_mount server_2 |grep ckpt
ckpt_test_ckpt_001 on /root_vdm_1/ckpt_test_ckpt_001 ckpt,perm,ro
ckpt1 on /ckpt1 ckpt,perm,ro
ckpt1_replica1 on /ckpt1_replica1 ckpt,perm,ro
then copy it back to the file system:
# server_archive server_2 -rw -i /ckpt1/RESTORE_ME.txt /FS_RAINER
ATTENTION: server_archive interactive file rename operation.
-rw-r--r-- Aug 5 10:42 RESTORE_ME.txt
Input new name, or a "." to keep the old name, or a "return" to skip this file.
Input > >>RESTORE_ME_COPY.txt
Processing continues, name changed to: RESTORE_ME_COPY.txt
# ls -l FS_RAINER
total 16
drwxr-xr-x 2 root root 8192 Aug 5 10:40 lost+found
-rw-r--r-- 1 root root 56 Aug 5 10:42 RESTORE_ME_COPY.txt
Without the -i server_archive will copy the file or the whole directory using the same names
Unfortunately there doesnt seem an easy way to specify a new destination file name when working
non-interactively (see -s option if you are familiar with ed Syntax)
you could also traverse the checkpoints using the names that are displayed in the .ckpt dir like this:
# server_archive server_2 -rw -i /FS_RAINER/.ckpt/2009_08_05_10.43.17_Europe_Berlin/RESTORE_ME.txt /FS_RAINER
ATTENTION: server_archive interactive file rename operation.
-rw-r--r-- Aug 5 10:42 RESTORE_ME.txt
Input new name, or a "." to keep the old name, or a "return" to skip this file.
Input > >>RESTORE_ME_NEW2.txt
Processing continues, name changed to: RESTORE_ME_NEW2.txt