SEO

Does anyone presently use the splitvg command with a volume group
containing JFS2 file systems? We used to use the chfs -a splitcopy on
the JFS file systems to separate a mirror and use the read-only copy to
make a backup.

We are looking at doing this at the VG level now but are curious to
know:

1) How long does it take for the splitvg to break its mirror on (for
example) two 72GB disks? (one disk plus the mirror disk)

2) How long does it take for the joinvg to join back up and be sync'd
again?

We are going to run a few tests as soon as we free up a server here,
but just was curious if anyone is using this now and any good/bad
experiences they have had with it.

TIA,

Steve

Remark: As JFS2 is used, a (newer) capability is snapshot. There is no
need to break a mirror and resync it.

Space is needed in the vg for the JFS2 changes during the lifetime of
the snapshot.

Bernard Dhooghe


steven_nospam at Yahoo! Canada wrote:
> Does anyone presently use the splitvg command with a volume group
> containing JFS2 file systems? We used to use the chfs -a splitcopy on
> the JFS file systems to separate a mirror and use the read-only copy to
> make a backup.
>
> We are looking at doing this at the VG level now but are curious to
> know:
>
> 1) How long does it take for the splitvg to break its mirror on (for
> example) two 72GB disks? (one disk plus the mirror disk)
>
> 2) How long does it take for the joinvg to join back up and be sync'd
> again?
>
> We are going to run a few tests as soon as we free up a server here,
> but just was curious if anyone is using this now and any good/bad
> experiences they have had with it.
>
> TIA,
>
> Steve

Bernard Dhooghe wrote:
> Remark: As JFS2 is used, a (newer) capability is snapshot. There is no
> need to break a mirror and resync it.
>
> Space is needed in the vg for the JFS2 changes during the lifetime of
> the snapshot.
>
> Bernard Dhooghe

Hi Bernard,

I tried snapshot a while back when I was first informed about it, but
the command did not work as planned. It creates images of a particular
file system, but these images did not appear to allow us to access the
files in the snapshot, it just let you see it as one large binary file.

What we want to do is create a backup tape on DDS4 tapes in either tar
or cpio format which contains a copy of the file system contents. We
have a datavg volume group, and it contains a series of file systems
listed under the /live mount point. The file systems contain ISAM data
files, COBOL programs, runtime software, and various reference files.

In the JFS world, we used to splitcopy the files system into a live and
frozen version, and we would then run our backups in the frozen version
while users would continue to work in the live version. In the JFS2
world, we want something similar to this. splitvg looks like it offers
the closest similar method.

"steven_nospam at Yahoo! Canada" <steven_nospam@yahoo.ca> wrote in message
news:1163193098.378157.230190@b28g2000cwb.googlegr oups.com...
> the command did not work as planned. It creates images of a particular
> file system, but these images did not appear to allow us to access the
> files in the snapshot, it just let you see it as one large binary file.

the 'image' is actually a lv of copy-on-write block data.. but its mountable
as a filesystem. e.g. assuming jfs2 volume /bob.

snapshot -o snapfrom=/bob -o size=1G # creates a snap lv typically named
/dev/fslv<xx>
mount -o snapshot /dev/fslv<xx> <snap_mountpoint> # say, mount -o snapshot
/dev/fslv00 /snap/bob
(do whatever you do to back up files in /snap/bob)
umount /snap/bob
snapshot -o remove /dev/fslv<xx>

i occasionally do this as a cya/backup measure for a volume holding mysql db
stuff.. works well.

-r

Remark 2: the subsystem (isam,database) must provide kind of checkpoint
or suspend I/O capability to have a backup that is fully consistent
after restore and optional recovery. The file system snapshot itself
can not provide this (neither can splitvg).

Bernard Dhooghe

Bernard Dhooghe wrote:
> Remark: As JFS2 is used, a (newer) capability is snapshot. There is no
> need to break a mirror and resync it.
>
> Space is needed in the vg for the JFS2 changes during the lifetime of
> the snapshot.
>
> Bernard Dhooghe
>
>
> steven_nospam at Yahoo! Canada wrote:
> > Does anyone presently use the splitvg command with a volume group
> > containing JFS2 file systems? We used to use the chfs -a splitcopy on
> > the JFS file systems to separate a mirror and use the read-only copy to
> > make a backup.
> >
> > We are looking at doing this at the VG level now but are curious to
> > know:
> >
> > 1) How long does it take for the splitvg to break its mirror on (for
> > example) two 72GB disks? (one disk plus the mirror disk)
> >
> > 2) How long does it take for the joinvg to join back up and be sync'd
> > again?
> >
> > We are going to run a few tests as soon as we free up a server here,
> > but just was curious if anyone is using this now and any good/bad
> > experiences they have had with it.
> >
> > TIA,
> >
> > Steve

No Body wrote:
> "steven_nospam at Yahoo! Canada" <steven_nospam@yahoo.ca> wrote in message
> news:1163193098.378157.230190@b28g2000cwb.googlegr oups.com...
> > the command did not work as planned. It creates images of a particular
> > file system, but these images did not appear to allow us to access the
> > files in the snapshot, it just let you see it as one large binary file.
>
> the 'image' is actually a lv of copy-on-write block data.. but its mountable
> as a filesystem. e.g. assuming jfs2 volume /bob.
>
> snapshot -o snapfrom=/bob -o size=1G # creates a snap lv typically named
> /dev/fslv<xx>
> mount -o snapshot /dev/fslv<xx> <snap_mountpoint> # say, mount -o snapshot
> /dev/fslv00 /snap/bob
> (do whatever you do to back up files in /snap/bob)
> umount /snap/bob
> snapshot -o remove /dev/fslv<xx>
>
> i occasionally do this as a cya/backup measure for a volume holding mysql db
> stuff.. works well.
>
> -r

Sorry to break in here.
I have been trying to get snapshot to work, and I think it is broken.
I am trying this on AIX 5.3 TL 5.03 (latest as of last week).

Using your example commands this is what I get:
# snapshot -o snapfrom=/c040 -o size=64M
Snapshot for file system /c040 created on /dev/fslv00
# mount -o snapshot /dev/fslv00 /tmp/snap
mount: invalid mount option: "snapshot"
mount: There is no plausible log device for /dev/fslv00 ("/tmp/snap")
#

What am I missing? Both the mount man page and the snapshot man page
seem to agree what I am doing should work.

Hi Steve,

We extensibely use the splitvg and joinvg options on our environment.
Bascially to have a replica sort of file system . Our database is file
based and around 100 GB in size. It's a lot depend on the changes to
the master copy , if you are attaching back ( joinvg ) .

Where as splitvg is instant and you can mount a file system over the
other copy.

Let me know if that is what you are asking .
Some thing to note, You need to create a lv first tjen create the FS
and extend the vg to the other set of hard disk.

After that make use of mklvcopy

Regrds
RC

steven_nospam at Yahoo! Canada wrote:
> Does anyone presently use the splitvg command with a volume group
> containing JFS2 file systems? We used to use the chfs -a splitcopy on
> the JFS file systems to separate a mirror and use the read-only copy to
> make a backup.
>
> We are looking at doing this at the VG level now but are curious to
> know:
>
> 1) How long does it take for the splitvg to break its mirror on (for
> example) two 72GB disks? (one disk plus the mirror disk)
>
> 2) How long does it take for the joinvg to join back up and be sync'd
> again?
>
> We are going to run a few tests as soon as we free up a server here,
> but just was curious if anyone is using this now and any good/bad
> experiences they have had with it.
>
> TIA,
>
> Steve

mount -v jfs -o snapshot /...
generates also the reported error.

So a solution (avoiding also any default that could change between MLs
or disappear with new capabilities) will probably be:

mount -v jfs2 -o snapshot /...

Bernard Dhooghe


demianphillips@gmail.com wrote:
> No Body wrote:
> > "steven_nospam at Yahoo! Canada" <steven_nospam@yahoo.ca> wrote in message
> > news:1163193098.378157.230190@b28g2000cwb.googlegr oups.com...
> > > the command did not work as planned. It creates images of a particular
> > > file system, but these images did not appear to allow us to access the
> > > files in the snapshot, it just let you see it as one large binary file.
> >
> > the 'image' is actually a lv of copy-on-write block data.. but its mountable
> > as a filesystem. e.g. assuming jfs2 volume /bob.
> >
> > snapshot -o snapfrom=/bob -o size=1G # creates a snap lv typically named
> > /dev/fslv<xx>
> > mount -o snapshot /dev/fslv<xx> <snap_mountpoint> # say, mount -o snapshot
> > /dev/fslv00 /snap/bob
> > (do whatever you do to back up files in /snap/bob)
> > umount /snap/bob
> > snapshot -o remove /dev/fslv<xx>
> >
> > i occasionally do this as a cya/backup measure for a volume holding mysql db
> > stuff.. works well.
> >
> > -r
>
> Sorry to break in here.
> I have been trying to get snapshot to work, and I think it is broken.
> I am trying this on AIX 5.3 TL 5.03 (latest as of last week).
>
> Using your example commands this is what I get:
> # snapshot -o snapfrom=/c040 -o size=64M
> Snapshot for file system /c040 created on /dev/fslv00
> # mount -o snapshot /dev/fslv00 /tmp/snap
> mount: invalid mount option: "snapshot"
> mount: There is no plausible log device for /dev/fslv00 ("/tmp/snap")
> #
>
> What am I missing? Both the mount man page and the snapshot man page
> seem to agree what I am doing should work.

Bernard Dhooghe wrote:
> mount -v jfs -o snapshot /...
> generates also the reported error.
>
> So a solution (avoiding also any default that could change between MLs
> or disappear with new capabilities) will probably be:
>
> mount -v jfs2 -o snapshot /...
>
> Bernard Dhooghe
>

Thank you!

I feel so stupid looking at that.