Need help with File Systems...

[franksoprano]

I am having trouble increasing my file system sizes.. when i run smitty fs and try to add additional space to the file system it returns with the following:

0516-404 allocp: This system cannot fulfill the allocation request.There are not enough free partitions or not enough physical volumes to keep strictness and satisfy allocation requests. The command should be retried with different allocation characteristics.

------------

here is my df -k:

Filesystem 1024-blocks Free %Used Iused %Iused Mounted on
/dev/hd4 2260992 1438452 37% 1371 1% /
/dev/hd2 786432 15872 98% 23997 13% /usr
/dev/hd9var 16384 13760 17% 220 6% /var
/dev/hd3 98304 44684 55% 3328 14% /tmp
/dev/hd1 966656 935808 4% 21 1% /home
/dev/datalv 1638400 1263988 23% 2232 1% /datafs

--------------


I have 2 4.5Gig drives in my system with netscape and base os thats it...

What do I have to do to increase the FS?

Thanks!

 

[CaptNeo]

Which FS are you trying to increase?

 

[franksoprano]

I would like to increase the /usr FS...

 

[nAtHaNs]

hi,

you don't need to increase /usr filesystem unless you are saving something in that directory which is not advisable. /usr is the system repository of all lpp's (licensed program products). It is not dynamic, therefore it does not increase or decrease. Above 90% used is just normal.

 

[franksoprano]

I am trying to install samba and it wants to place some files in the /usr fs... it returns:

0503-008 installp: there is not enough free disk space in filesystem /usr (34666 more 512-byte blocks are required.) An attempt to extend this filesystem was unsuccessful. Make more space available then retry this option

 

[AIX5L]

Look in your Volumegroup of Available Space:
lsvg rootvg or on your Disk lspv hdiskn
You can only increase the Size if you have Space in the VG.
I prefer to put /datafs in other VGs then the rootvg.

 

[franksoprano]

here is my lsvg rootvg:

VOLUME GROUP: rootvg VG IDENTIFIER: 00ffffff35398d49
VG STATE: active PP SIZE: 16 megabyte(s)
VG PERMISSION: read/write TOTAL PPs: 270 (4320 megabytes)
MAX LVs: 256 FREE PPs: 0 (0 megabytes)
LVs: 8 USED PPs: 270 (4320 megabytes)
OPEN LVs: 7 QUORUM: 2
TOTAL PVs: 1 VG DESCRIPTORS: 2
STALE PVs: 0 STALE PPs: 0
ACTIVE PVs: 1 AUTO ON: yes
MAX PPs per PV: 1016 MAX PVs: 32


here is the lspv hdisk0:

PHYSICAL VOLUME: hdisk0 VOLUME GROUP: rootvg
PV IDENTIFIER: 00ffffff525a2148 VG IDENTIFIER 00ffffff35398d49
PV STATE: active
STALE PARTITIONS: 0 ALLOCATABLE: yes
PP SIZE: 16 megabyte(s) LOGICAL VOLUMES: 8
TOTAL PPs: 270 (4320 megabytes) VG DESCRIPTORS: 2
FREE PPs: 0 (0 megabytes)
USED PPs: 270 (4320 megabytes)
FREE DISTRIBUTION: 00..00..00..00..00
USED DISTRIBUTION: 54..54..54..54..54

lspv hdisk1:
PHYSICAL VOLUME: hdisk1 VOLUME GROUP: datavg
PV IDENTIFIER: 00ffffffcfc697ff VG IDENTIFIER 00ffffff39d9ba70
PV STATE: active
STALE PARTITIONS: 0 ALLOCATABLE: yes
PP SIZE: 8 megabyte(s) LOGICAL VOLUMES: 2
TOTAL PPs: 518 (4144 megabytes) VG DESCRIPTORS: 2
FREE PPs: 317 (2536 megabytes)
USED PPs: 201 (1608 megabytes)
FREE DISTRIBUTION: 103..00..07..103..104
USED DISTRIBUTION: 01..104..96..00..00

 

[klasan]

Your rootvg volume group is 100% used, you cannot increase /usr fs.
If /home fs is empty remove and recreate it smaller..
umount /home
rmfs -r /home
mklv -y hd1 rootvg 2
crfs -v jfs -d hd1 -m /home -A y
Now you can increase the size of /usr:
chfs -a size=+50000 /usr

Your / fs is 2.2 Gb with 1.4Gb free space....bad idea...

n.b. sorry for my english....i'm Italian...

 

[AIXSPadmin]

In posting to your original question, your error is indicating that you don't have enough logical partitions in your logical volume. Increase the number of logical partitions for hd2, then you can increase your filesystem size.

Example:
If your logical volume lvx1 is in datavg which has 16MB partitions, and lvx1 is 10 partitions. This will make your filesystem /newfilesystem 160MB. You also have 10 LPs for lvx1. Now you want to increase /newfilesystem by 160MB, doubling it, before you can do that you need to add 10 LPs to lvx1 giving you 20 LPs and then you can add 160MB to /newfilesystem.

 

[Yegolev]

Your / fs is way, way too big. Just checking one of my machines (most of which are very large systems, if you have not been reading my drivel):

me@myhost:/utc/home/me>df -k
Filesystem 1024-blocks Free %Used Iused %Iused Mounted on
/dev/hd4 131072 103204 22% 2905 5% /
/dev/hd2 1572864 395832 75% 44740 12% /usr
/dev/hd9var 196608 170980 14% 521 2% /var
/dev/hd3 131072 117116 11% 222 1% /tmp
/dev/hd1 131072 119768 9% 370 2% /home
...

There are more fs, but you see that I'm only using ~30MB of /. Also note that my /usr is ~1.5GB. I'd suggest remaking your system, if possible, using this as a guide. IBM Certified -- AIX 4.3 Obfuscation

 

[aixmurderer]

Quite interesting to what all have to say about putting what where.
At the previous company I worked for all applications lived in /home, that's the way I got to know FS usage.

At my present company (as set by IBM originally I think, we are a Business partner and equipment was leased from and configured/maintained by IBM, years before my arrival) everything was installed (DB2, application software, etc.) in /usr/local/ (/usr FS).

Now in our case DB2 isn't too bad as it's a reasonably static DB used for billing only (field number incrementing), but as the system grew the /usr had to be extended to accomodate a lot more applications and file archiving (we do EDI mapping). IBM documentation seems to favour /home for creating the DB instances.

What is regarded as the accepted standard? I have noticed Linux tends to install a lot of stuff in /usr/local (e.g Apache), but then lately the document root in certain cases (RH7.x) for Apache has moved to /var/www. Some earlier documentation I have seen recommend to have the document root living in /home. There just seems to be no accepted standard, or am I just ignorant here?

Quite confusing...



IBM Certified Confused - MQSeries
IBM Certified Flabbergasted - AIX 5 pSeries System Administration

 

[MoshiachNow]

I would first check for the biggest directories in /usr,just in case something made it's way there somehow unintentionly:

du -a /usr | grep -v .rsrc | sort -n -r | head -10
"Long live king Moshiach !"
h

http://www.noahide.com/movement.htm

 

[MoshiachNow]

I would first check for the biggest directories in /usr,just in case something made it's way there somehow unintentionly:

du -a /usr | sort -n -r | head -10
"Long live king Moshiach !"
h

http://www.noahide.com/movement.htm

 

[mrn]

Create a new FS called /usr/local/samba and stick it in there.

--
| Mike Nixon
| Unix Admin
|

http://www.instantflashshop.com

----------------------------

 

[Yegolev]

I do not know what the standard is, but I understand the reason why we do things the way we do.

First, most of the machines are "large" and have at least Oracle, but not too many other apps. The datafiles each get their own fs. There is also /oracle and /oracle/SID on the SAP machines, and the SAP binaries all get their own fs, plus separate fs for redo logs, etc. All that stuff goes in the places the vendor's installer puts things (afaik).

For stuff like Perl and Apache, we just stick it where the installer puts it, unless it is a piece of bloatware. We try to give it just enough room to breathe, especially if no growth is expected. We try to leave over 10% free so that the fs monitor does not alarm, if we can.

We do not mksysb to tape, so the nightly mksysb is sent as an image to its own fs, then archived via TSM. Naturally we would want rootvg to be as small as possible, and in most cases rootvg is the only vg on internal disks. Sometimes there's a internal pagingvg, depending on how the hardware is set up.

Of course we have disk coming out our ears (sort of), so we can do this sort of thing. IBM Certified -- AIX 4.3 Obfuscation

 

[bi]

aixmurderer,

I learned Unix in an HP shop with lots of excellent Unix sys admins. Then I moved to an AIX shop and they had pretty much the same standards. This is how I learned to lay out FS:

/home is for the users to put their data in, not applications. On development systems, /home may become quite large, depending on how the development team does its work. On production systems, /home usually isn't that large.

Applications go in /usr (for AIX), except for apps like Oracle, which always has its own FS. /usr is the default location when you install apps on AIX and is the largest of the default filesystems because of that.

/usr/local is for shareware programs like Samba and should be a separate filesystem (to prevent overwriting during an OS upgrade).

/ should always contain nothing but what the OS wants there (etc, dev, etc.) and is usually the smallest of all the default filesystems.

And then, you have other filesystems for the database files, which allow you to spread out your filesystems across disks and controllers for performance and data protection.

 

[aixmurderer]

Bi - sounds good to me, but now for the curve ball, where does /opt fit into the whole scenario?

On AIX the freeware GNU stuff ends up there, on Solaris and Linux some software like MQSeries uses /opt as well as /var. Now the /var bit for MQ queues I can understand as they will vary according to depth, but why the software in /opt instead of /usr?

IBM Certified Confused - MQSeries
IBM Certified Flabbergasted - AIX 5 pSeries System Administration

 

[bi]

aixmurder,

Not such a big curve ball.

With 5L, I think /opt is created by default (right?). And, of course, for Solaris and HP parts of apps (or all of some apps) are put into /opt. So /opt is a default OS fileystem on those systems.

Each software has its own defaults, it seems, on where it wants its various files to be placed. Whether it puts most of everything in /usr or /opt probably depends on where the application came from. Was it written exclusively for AIX 4.x or ported from Linux development for use on 5L, for example. If it was first written for 4.x, I would expect to see everything in /usr, even on 5L. If it was originally written on another platform and ported over, I can see where you might find files in /opt instead.

And then you have the applications that have files all over the place. For example, on HP, OmniBack has important files in /etc/opt/omni (making it part of the / filesystem), /opt/omni, and /var/opt/omni.

Back to AIX, version 4.3: All of my freeware GNU stuff I put in /usr/local. That way, I know freeware/shareware will always be in /usr/local. /var will always have a lot of "stray" files for logs and such.

I think there are a couple of points or standards that I see eroding: people dumping everything into plain old directories in / because they probably got their start in Linux (which doesn't seem to have the concept of filesystems and how to use LVM for performance and data protection -- I've seen Linux, but haven't been a sys admin for it) or they got their start with Solaris on Intel, and the idea of keeping the entire rootvg lean and mean so you can get your OS back quickly in the case of a disaster or mini-disaster. (It takes a lot longer for a mksysb to get you back on your feet if it is bloated with stuff that probably should be in another volume group. I've even created a /home/WebSphere filesystem for the home directories of our WebSphere development group and put it in another volume group just to avoid having a bloated rootvg.)

This isn't the first post I've seen from folks having problems with rootvg and when they post their df -k, more often than not, / is 2 or 3 GB. And not using LVM to balance the load across controllers and disks is going to start giving Unix (whether it's AIX, HP or Solaris) a bad name when databases grind to a halt because the disks and controllers aren't properly load balanced. And it's so easy to do with AIX, even if you only have a few disks.

Sorry to have gone on for so long.

 

[aixmurderer]

bi - and excellent discussion, we should perhaps take this somewhere else, perhaps even a FAQ.

I think I'm not the only one out there who get confused with all these FS, especially when you are involved in more than AIX. In my case it's AIX, Linux, Solaris and SCO, to try and achieve some sort of standard across the board is difficult.

Then we still have DB2, which unlike Oracle and Cincom Supra DB(my other exposure) doesn't live in raw space, but in rootvg or another user defined vg.

My predicament at this moment is, should I mirror rootvg across the 3 disks + 1 hot spare I have (done so), or should I stripe for performance? We don't have HACMP, so availability is important, so cannot afford to lose one disk in a quorum. Perhaps I should stripe over 2 disks and mirror to the other 2, with the bootlist then listing hdisk0 and hdisk2 as boot disks. Not sure if this will work though...

IBM Certified Confused - MQSeries
IBM Certified Flabbergasted - AIX 5 pSeries System Administration

 

[bi]

Don't worry about quorum in a mirrored VG. When I mirrored my rootvg (there are only two disks in my rootvg), I believe the mirrorvg rootvg command automatically turned off the quorum requirement -- at least I don't remember turning it off myself. (I think rootvg is the only VG that quorum is automatically turned off.)

After you mirror any VG, I'd double-check that quorum is turned off. It would really be frustrating to have a 2-disk mirrored volume group fail because of quorum.

I'm not too familiar with striping and mirroring. If you have older disks, I can see where that might help performance. But if you have disks that spin at 10K, I wouldn't worry too much about striping -- unless, of course, you have files from a database in rootvg. (tsk, tsk!)

I'd like to write some FAQs, but there are so many more folks here who know so much more than me.

Thanks.