Features of Solaris 2.4 journaled file system ?

Features of Solaris 2.4 journaled file system ?

Post by Ole Holm Niels » Tue, 10 Jan 1995 17:29:37



I have read that Solaris 2.4 comes with a "journaled file system",
but we're still waiting for our support contract Solaris 2.4 to arrive.

Can anyone explain the main features of Solaris 2.4 journaled file
system ?  In particular, I'd be interested to find out if Sun added
volume management so that one may distribute a file system over
multiple physical disks and even enlarge filesystems on the fly.
We have come to appreciate this feature very much on our AIX 3.2
machines, and I wonder when SunOS will follow.

Ole H. Nielsen
Department of Physics, Building 307
Technical University of Denmark, DK-2800 Lyngby, Denmark

Telephone: (+45) 45 93 12 22 - 3187
Telefax:   (+45) 45 93 23 99

 
 
 

Features of Solaris 2.4 journaled file system ?

Post by Casper H.S. D » Tue, 10 Jan 1995 18:37:03



Quote:>I have read that Solaris 2.4 comes with a "journaled file system",
>but we're still waiting for our support contract Solaris 2.4 to arrive.
>Can anyone explain the main features of Solaris 2.4 journaled file
>system ?  In particular, I'd be interested to find out if Sun added
>volume management so that one may distribute a file system over
>multiple physical disks and even enlarge filesystems on the fly.
>We have come to appreciate this feature very much on our AIX 3.2
>machines, and I wonder when SunOS will follow.

Solaris 2.4 comes in several different packagings.
The server versions (workgroup & enterprise server) come bundled
with a product called "Online: Disksuite 3.0".
The desktop version doesn't come with OLD.

The baseos supports ufs logging with hooks
left & right, but doesn't implement it.

Online Disksuite implements logging.

Logging has nothing to do with growing filesystem.
That is another feature of ODS.

With ODS you can grow filesystems on metadevices.  A metadevice
is a set of one or more normal file partitions that are made to
look like one partition.  A metadevice can be increased in size
by adding real partitions, the filesystem on a meta device can then
be grown with growfs.

Casper

 
 
 

Features of Solaris 2.4 journaled file system ?

Post by sc.. » Wed, 11 Jan 1995 23:51:53


Hi,

I have added what I've been sent as entries to cmtp.conf (sp? sorry I'm sending this
from a non-Solaris OS at present) and still see an 8500 when I do mt status.

Can I change this? Is it in fact getting it all wrong?

I have managed to both read and write tars and dumps so far, so I am at least
talking to the drive.

Any ideas?

Thanks

Derek

--------------------------------------------------------------------------------
Derek Jones                                          !SCUNA Computer Consultancy
Computer Consultant                                  ! 53 Archer Road, Sheffield
SCUNA (Systems,C,UNIX,Networking,Advice)             !                    S8 0JT

WWW: http://www.veryComputer.com/!   Tel: +44 (0)114 2555524
--------------------------------------------------------------------------------
"You're from **space*?" "No I'm from Sheffield, I only *work* in *space"

 
 
 

Features of Solaris 2.4 journaled file system ?

Post by Ian Dickins » Thu, 12 Jan 1995 01:18:36




Quote:>1. Does this mean that we still have the usual <= 8 disk partitions
>per physical disk, and that such partitions can be concatenated
>to enlarge the metadevice ?

Yes.

Quote:>The AIX 3.2 ability to add disk chunks
>(PPs) of typically 4 MB at a time then seems rather nicer.

Possibly.

Quote:>2. If one starts out creating a normal filesystem, can it later on be
>transformed into a metadevice, in case one finds that more disk space is
>needed ?

Not currently.

Quote:>3. I take it that filesystems > 2 GB are possible with OnLine:Disksuite,
>right ?  What is the limit actually, will we ever hit it ?

Yes they are possible, though I can't remember what the limit is.

Quote:>4. Can growfs be run while the filesystem is mounted, or must the filesystem
>be taken out of service while running growfs ?

We've used growfs on homedir and usenet metadisks whilst up and running.
It works like magic.

Cheers,
--
Ian 'Vato'*inson [ID17]                                   Kibo bait :-)

           MIME mail welcome - don't send me no s*kin' X.400
      Click <A HREF="http://www.veryComputer.com/~cudep/">here</A>.

 
 
 

Features of Solaris 2.4 journaled file system ?

Post by Frank Pete » Fri, 13 Jan 1995 08:59:29




>> growfs can be run while the filesystem is mounted.
>> The filesystem will be write locked in the process,
>> so people writing to the disk will notice.

>did You test this or do You just trust the manual.

I've actually done it (at least three times that I can think of).  It
works fine.  As Casper indicated, writes pause while the growfs is
running (which isn't very long...it took a few seconds to add a stripe
consisting of two 2.1GB differential SCSI disks to a metadevice
consisting of three similar stripes).

We've had OLD since it was shipped as SPARCserver Manager.  We've run
it on several fairly large (thousands of userids, tens of gigabytes)
systems and it has always performed as advertised.
--
Frank Peters  -  UNIX Systems Group Leader  -  Mississippi State University

             WWW Home Page:  http://www.msstate.edu/~fwp/

 
 
 

Features of Solaris 2.4 journaled file system ?

Post by Ole Holm Niels » Fri, 13 Jan 1995 18:05:13




> >1. Does this mean that we still have the usual <= 8 disk partitions
> >per physical disk, and that such partitions can be concatenated
> >to enlarge the metadevice ?  The AIX 3.2 ability to add disk chunks
> >(PPs) of typically 4 MB at a time then seems rather nicer.
> I'm not familiar how this works in AIX.  I don't see why AIX has a particular
> advantage here, growing a filesystem doesn't look like something you'd
> want to do in such small icnrements, causing your filesystem to be
> shattered all over the place.

Well, the AIX filesystem is pretty neat on this point:  You can enlarge
filesystems down to the PP unit (4 MB).  If your filesystem gets spread
over the disk, no sweat: You can simply rearrange the filesystems while
they're *live* and *mounted*: You invoke the /etc/reorgvg command, which
will move blocks around according to a user-specified priority !
You typically put paging space (swapspace) in the center, and less used
file systems in the peripheries.

I can think of several occasions where I wished my SunOS /usr partition
had been created, say, 10 MB larger, like when adding yet another
software package.  So adding small increments to filesystems is most
convenient for the system administrator.

Quote:> >4. Can growfs be run while the filesystem is mounted, or must the filesystem
> >be taken out of service while running growfs ?
> growfs can be run while the filesystem is mounted.
> The filesystem will be write locked in the process,
> so people writing to the disk will notice.

Another point where the AIX filesystem wins :-)

Ole H. Nielsen
Department of Physics, Building 307
Technical University of Denmark, DK-2800 Lyngby, Denmark

Telephone: (+45) 45 93 12 22 - 3187
Telefax:   (+45) 45 93 23 99

 
 
 

Features of Solaris 2.4 journaled file system ?

Post by Bruce Adler [Contracto » Sat, 14 Jan 1995 06:12:03




>Hi,

>I have added what I've been sent as entries to cmtp.conf (sp?  sorry I'm

I may have missed the beginning of this thread so I'm not exactly clear
what you're tyring to accomplish.

Who sent you those cmtp.conf updates?  I don't know where the idea
originated that it's possible to add entries to cmtp.conf for new tape
drives on Solaris x86 .  That's a feature of the SPARC st driver not the
x86 cmtp or sctp drivers.  There currently isn't any method for a user
to add a new tape drive to cmtp (or sctp).  Currently, the table of
known drives is compiled into sctp and can only be modified via a Driver
Update release.  

Quote:>sending this from a non-Solaris OS at present) and still see an 8500
>when I do mt status.  

The mt program has a bug. It doesn't accurately report the drive type.
If you hooked up the drive and did a reconfigure reboot the drive should
be usable. If it's not a support model you won't be able to use any
of its "optional" features.

Quote:>Can I change this? Is it in fact getting it all wrong?

Mt is getting it wrong. Sorry, you can't change mt, cmtp or sctp.

Quote:>I have managed to both read and write tars and dumps so far, so I am at
>least talking to the drive.  

All SCSI-2 compatable drives should work in their default states/modes
without adding their configurations to the Solaris x86 sctp driver.  You
only have to add the tape drive to the configuration table if you
need/want to use any of the optional SCSI-2 features such as variable
length records, backspace record/file, alternate densities, unload the
tape, etc.  The default configuration is sort of like the very old
Archive Viper QIC-150 drive.  

Quote:>Any ideas?  

Use a tape drive which is on the Solaris x86 2.4 Hardware Compatibility
List or wait for the Driver Update that adds whatever drive you're
trying to use.  

BA

 
 
 

Features of Solaris 2.4 journaled file system ?

Post by Rob McMah » Fri, 13 Jan 1995 21:51:42



>>> growfs can be run while the filesystem is mounted.  The filesystem will be
>>> write locked in the process, so people writing to the disk will notice.

>>did You test this or do You just trust the manual.

>I've actually done it (at least three times that I can think of).  It works
>fine.

I'd just like to confirm this.  We've been running Disksuite for about 2?
years now and I've grown filesystems online on several occasions, not just by
adding metadevices to existing concatenations, but by re-partitioning disks
and growing (non-metadevice) filesystems into the new partition sizes.  I've
never had a problem with Disksuite.

Cheers,

Rob
--
UUCP:   ...!mcsun!uknet!warwick!cudcv   PHONE:  +44 203 523037

Rob McMahon, Computing Services, Warwick University, Coventry CV4 7AL, England

 
 
 

Features of Solaris 2.4 journaled file system ?

Post by J.A. Pete » Sat, 21 Jan 1995 09:17:07



Quote:> Well, the AIX filesystem is pretty neat on this point:  You can enlarge
> filesystems down to the PP unit (4 MB).  If your filesystem gets spread
> over the disk, no sweat: You can simply rearrange the filesystems while
> they're *live* and *mounted*: You invoke the /etc/reorgvg command, which
> will move blocks around according to a user-specified priority !
> You typically put paging space (swapspace) in the center, and less used
> file systems in the peripheries.

> I can think of several occasions where I wished my SunOS /usr partition
> had been created, say, 10 MB larger, like when adding yet another
> software package.  So adding small increments to filesystems is most
> convenient for the system administrator.

The AIX jfs is pretty slick but you can get a reasonable facsimile out
of SunOS 4.1.x using the following scheme:

Here is quick overview of the current CAS filesystem layout with
respect to /usr/local & /local.

/local serves as the mount point for all external application disks.
all filesystems are mounted as /local/u{$FS_#} for managing storage
given the prevailing 2G filesystem limitation, ie:

sun%df -t 4.2
Filesystem            kbytes    used   avail capacity  Mounted on
/dev/sd1g            1714984 1467928   75558    95%    /local/u1
/dev/sd1h            1728040 1418714  136522    91%    /local/u2
/dev/sd3g            1714984 1119391  424095    73%    /local/u3
/dev/sd3h            1714984 1355590  187896    88%    /local/u4
/dev/sd2g            1921030 1502863  226064    87%    /local/u5
/dev/sd2h            1921030 1541669  187258    89%    /local/u6

ibm%lsvg
rootvg
localvg
ibm%%lsvg -l localvg
localvg:
LV NAME             TYPE       LPs   PPs  PVs  LV STATE      MOUNT POINT
u1                  jfs        400   400  1    open/syncd    /local/u1
u2                  jfs        400   400  1    open/syncd    /local/u2
loglv00             jfslog     1     1    1    open/syncd    N/A

a vendor/product directory tree is symbolic linked to the filesystem ie:

sun% ls -lg /local/tban
lrwxrwxrwx  1 root     petro           7 Oct 10 11:29 /local/tban -> u2/tban

the version directory is made in vendor/product directory ie:

sun% cd /local/tban
sun% ls -lg
total 1
drwxr-xr-x  3 ajelsner petro         512 Aug  3 17:50 2.1

/usr/local serves as the revision and location control for applications.
this is where symbolic links select /local, locally installed applicatons,
or /tool, automounted networked applications.

sun% %ls -lg /usr/local/tban
lrwxrwxrwx  1 mkmcleod petro          20 Oct 10 11:29 /usr/local/tban ->
/local/tban/2.1/sun4

sun%ls -lg /usr/local/XDesigner
lrwxrwxrwx  1 ajelsner vue            15 Sep 21 17:40 /usr/local/XDesigner ->
/tool/XDesigner

summary of the key points are:
 /local manages local storage of applications.
 /usr/local manages location and version of applications.

Cheers,