Volume Manager - Striping - sequence writing performance

Typical setup of a Oracle database server could be:
Systemdisk mirrored.
10 disks striped and mirrored as RAID 0+1 for the database-volume.
I have used the command: "time mkfile 1000m testfile" to test the sequence write performance. I wonder why I get performance of c.a. 60 sek when I write to the systemdisk (one mirrored disk), but only c.a. 38 sek when I write to the database-volume (5 striped disks mirrored with 5 other) !?
Why don't I get c.a. 4-5 times performance increase (c.a. 13-15 sek) ??
Do you have the same sindrome on your servers ?
Is this normal and then why ?
Regards
Halldor Gretar
p.s.
This sindrome is on all servers that I have tested, both setup by me and others, and both with Veritas Volume Manager and Sun DiskSuite.

Hello,
I�m not an expert on volume management, but maybe these considerations that come to my mind can help you to improve your performance:
1.- The interlace size of the striping. You should adjust the size of the striping to match the I/O requests made by the Operating System or by the database management software (is the data access in a raw mode?). For example, if the data access is made through normal ufs access, the stripping size should match the block size of the file system.
2.- Are those disks on different controlers? Maybe a saturation of the controler, of the bus, etc... could slow down your I/O read/writes.
Bye,
jmiturbe

Similar Messages

Poor Veritas Volume Manager performance?

I apologize if this is a basic question.
Environment:
Solaris 8
Veritas Volume Manager 3.0.4
Oracle 9iR2
Based on Oracle's recommendation, I turned on the mincache=direct,convosync=direct,largefiles options for the file system Oracle will reside. Then I install Oracle 9iR2. I notice the installation process takes considerably longer than before when I installed it on the other identical machine(E250, dual CPUs, 1280M RAM, RAID 5 with no logging under Vxvm).
Other than the mounting options shown above, I don't see any difference as far as Vxvm is concerned. The other fine system has option: mincache=closesync,largefiles set for the file system.
Can someone shed some lights on this as to how to identify what went wrong?
Appreciated!

Jeffrey,
thanks for the info.
Here is the output of my iostat -En: (they, including the boot disk, all have one Soft Error/Illegal Request)
# iostat -En
c0t0d0 Soft Errors: 1 Hard Errors: 0 Transport Errors: 0
Vendor: SEAGATE Product: ST318404LSUN18G Revision: 4207 Serial No: 3BT0XZ3J000
02105
Size: 18.11GB <18110967808 bytes>
Media Error: 0 Device Not Ready: 0 No Device: 0 Recoverable: 0
Illegal Request: 1 Predictive Failure Analysis: 0
c0t6d0 Soft Errors: 0 Hard Errors: 0 Transport Errors: 0
Vendor: TOSHIBA Product: XM6201TASUN32XCD Revision: 1103 Serial No: 12/12/97
Size: 18446744073.71GB <-1 bytes>
Media Error: 0 Device Not Ready: 0 No Device: 0 Recoverable: 0
Illegal Request: 0 Predictive Failure Analysis: 0
c0t8d0 Soft Errors: 1 Hard Errors: 0 Transport Errors: 0
Vendor: SEAGATE Product: ST318404LSUN18G Revision: 4207 Serial No: 3BT10VP9000
02109
Size: 18.11GB <18110967808 bytes>
Media Error: 0 Device Not Ready: 0 No Device: 0 Recoverable: 0
Illegal Request: 1 Predictive Failure Analysis: 0
c0t9d0 Soft Errors: 1 Hard Errors: 0 Transport Errors: 0
Vendor: SEAGATE Product: ST318404LSUN18G Revision: 4207 Serial No: 3BT10ZNG000
07108
Size: 18.11GB <18110967808 bytes>
Media Error: 0 Device Not Ready: 0 No Device: 0 Recoverable: 0
Illegal Request: 1 Predictive Failure Analysis: 0
c0t10d0 Soft Errors: 1 Hard Errors: 0 Transport Errors: 0
Vendor: FUJITSU Product: MAJ3182M SUN18G Revision: 0804 Serial No: 01P08417
Size: 18.11GB <18110967808 bytes>
Media Error: 0 Device Not Ready: 0 No Device: 0 Recoverable: 0
Illegal Request: 1 Predictive Failure Analysis: 0
c0t11d0 Soft Errors: 1 Hard Errors: 0 Transport Errors: 0
Vendor: FUJITSU Product: MAJ3182M SUN18G Revision: 0804 Serial No: 01P13137
Size: 18.11GB <18110967808 bytes>
Media Error: 0 Device Not Ready: 0 No Device: 0 Recoverable: 0
Illegal Request: 1 Predictive Failure Analysis: 0
c0t12d0 Soft Errors: 1 Hard Errors: 0 Transport Errors: 0
Vendor: SEAGATE Product: ST318404LSUN18G Revision: 8507 Serial No: 3BT0KZRQ000
07051
Size: 18.11GB <18110967808 bytes>
Media Error: 0 Device Not Ready: 0 No Device: 0 Recoverable: 0
Illegal Request: 1 Predictive Failure Analysis: 0

SVM equivalent command for veritas volume manager "VXEVAC" command

Hi All
I am working on a major migration project , where servers are heterogenous with part of servers with Veritas volume manager and rest with Solaris volume manager.
Migration is quite easy on Veritas servers using "VXEVAC COMMAND " i can easily move my data to new luns
But need to know any equivalent procedure in SVM. ..
all servers with latest solaris 10
Quick reply is highly appreciated.
Rgds
Md

Hello,
I�m not an expert on volume management, but maybe these considerations that come to my mind can help you to improve your performance:
1.- The interlace size of the striping. You should adjust the size of the striping to match the I/O requests made by the Operating System or by the database management software (is the data access in a raw mode?). For example, if the data access is made through normal ufs access, the stripping size should match the block size of the file system.
2.- Are those disks on different controlers? Maybe a saturation of the controler, of the bus, etc... could slow down your I/O read/writes.
Bye,
jmiturbe

Can't use built in volume manager in PCMan File Manager

Trying to use the built in volume manager in PCMan FM causes 2 error messages to appear. They are unreadable (the title bar says "Unable t" that's it). An error like the one that follows appears in the the console (or whatever Eterm is) for each:
(pcmanfm:3084): Gtk-WARNING **: Failed to set text from markup due to error parsing markup: Error on line 1 char 57: '-' is not a valid character following a '<' character; it may not begin an element name
After those a third message box appears and says: "Directory doesn't exist". That box is not accompanied by any error text on the console.
I assume the first to say something like "Unable to mount" but I don't know WHY.
The problem does NOT occur if I mount it as root (using sudo) and access it that way. This however then requires any further access to be made as root if anything useful is to be done with the drive (copying files, writing to it, etc.)
I'm running Fluxbox as my WM, if that matters.
Thanks in advance!

Hmm...the wiki refers to files I don't have. /etc/hal/fdi/policy and /etc/hal/fdi/information are both empty directories (even when viewed as root). Any thoughts?
I feel like an idiot now. I should have guessed i needed to put those files in.
Last edited by guitarMan666 (2009-04-22 03:22:36)

Linux LVM (Logical Volume Manager) for CentOS on Azure?

Hi. I am trying out Azure and installed a OpenLogic CentOS 6 virtual machine. I note that it is not running LVM (Logical Volume Manager) by default. I would like to ask if it is possible to:
1. have CentOS Linux installed with LVM by default when creating a Linux virtual machine on Azure
2. switch to LVM after adding a new disk
On the other hand, is it a good idea to use LVM at all? Will it affect performance, features on Azure?
Thanks.

Hi,
Based on my experience, you can add disk to an Azure VM. You can install the Logical Volume Manager to manage the disks attached to the VM. In addition, there is no Linux VM with LVM installed by default. If you want to have this, please submit your requirement
in Azure feedback:
http://feedback.azure.com/forums/34192--general-feedback
In addition, since you can have only one OS system disk for an Azure VM, this limitation may make multi-disk logical volume manager setups unworkable.
Best regards,
Susie
Please remember to mark the replies as answers if they help and unmark them if they provide no help. If you have feedback for TechNet Subscriber Support, contact [email protected]

How Solaris Volume Manager sync submirrors

HI Gurus,
I have a question on how Solaris volume manager (SVM) does re-synchronization (in raid 1). In another word, in case of one submirror was modified during boot process, how SVM detects it and how SVM validate which sumirror is gold.
One scenario I ran into: We had a software installed, this software updated /etc/name_to_sysnum which conflicts with new Solaris 10 release. So the system could not boot any more (not even to single user mode) after software installed. This box had root disk mirrored. To fix this, we boot from CDROM and mounted 1st mirror drive root partition (c0t0d0s0) and remove the bad entry in /etc/system (we did not break the mirror to make the changes). Then the box was able to boot up. After server was up, it was found /etc/system was rolled back with bad entries. Apparently it was synced back from 2nd mirror. So now the question is how SVM decides which submirror is invalid and should be re-sync from good submirror?
2nd scenario I saw: someone accidently added one root file system submirror into zone as a file system. But during zone installation, system got panic and rebooted. During reboot, system kept crashing. We managed to boot from network, break the mirror (update /etc/vfstab and /etc/system) and finally was able to boot the system. So in case one submirror was accidently accessed, how does SVM protect data and will the corrupted data written to disk slice synced to good submirror?
Please share your thoughts and point me with some good references. I could not find related info in SVM doc.
Thanks,
Wei

SVM doesnt "sync" disks ie copy data from one disk to another except in the case when your first setting up a mirror. Or your replacing a disk etc.
Those are circumstances when it realises the disks are out of sync.
Once it has a mirrored pair, it will keep them in sync since all writes will go to both sides.
And reads take alternate blocks from both disks.
So if the the two sides of a mirror have gotten out of sync, you will see strange results as half your content will come from one and half your content will come from the other. Even inside a single file, assuming the file is bigger than the stripe size.
So anything writing to one side of the mirror outside of SVM's control will corrupt things. And SVM has no mechanisms for detecting this and "fixing" things up.
So its vital to break the mirror if your going to be writing to the disk outside of SVM control.
If your brave and the amount of changes is small, you can try to edit both sides of the mirror.
But you have to remember that SVM works at the block level, not the filesystem level.
So if you do anything to make the two sides even minorly different. Even something as miror as update 2 files in a different order.
Then the layout of blocks on the disks in the 2 halves could be different. And your're screwed.
So don't do it on any system you care about. Its really easy to make a mistake and the consequences are usually catastrophic.
When in doubt break the mirrors. Its the only safe way.

Solaris Volume manager

Hi,
Hi,
We are in process of developing a backup utility (CLI) using the Raid manager (Firmware operating at device level).
Requirement
When the user gives a mount point for backup
-     Get the disk information that is mounted at the given mount point, if the mount point is for a metadevice then get all the disks under that metadevice.
-     Execute the Raid manager command for disk level copying for the disks that are mounted (Primary disk) to the other disk (secondary disk), which also can be under a metadevice.
-     Assign a backup ID, used for restoring the same when user wants
Problem
When a disk level copy operation is performed from a primary volume to a secondary volume which are configured by SVM, is there any way in which it is possible to configure the secondary (which now has the same data of primary) under SVM without any loss of data?

A new feature of Solaris Volume Manager will provide the ability to import DiskSets. It should be available in a latter update release of Solaris 9.

How to break Veritas Volume Manager mirrors and wipe disks

I have an old Solaris 10 server that is ready to be decommissioned. I would like to go in and break the Veritas Volume Manager mirrors and still be able to go in and wipe each of the two internal disks by using dd or something similar. Below is my Veritas Vol. Mgr. layout. I would greatly appreciate it if someone can post the exact commands (both for the Veritas Vol. Mgr. mirror break and the dd) I would need to run to perform the above tasks. I want to wipe the disks clean as to remove any company proprietary information prior to releasing the system.
Note: I will be logged into a remote console session to perform these actions as the system is in a datacenter and is located in a rack which does not have a monitor.
vxdisk list*
DEVICE TYPE DISK GROUP STATUS
c0t0d0s2 sliced rootdisk rootdg online
c0t1d0s2 sliced rootmirror rootdg online
vxdg list_
NAME STATE
rootdg enabled
vxprint -htr_
Disk group: rootdg
dg rootdg default default 0 1113432795.1025.dh22020
dm rootdisk c0t0d0s2 sliced 2632 71121402 -
dm rootmirror c0t1d0s2 sliced 2888 71124291 -
v appsvol - ENABLED ACTIVE 27262976 SELECT - fsgen
pl appsvol-01 appsvol ENABLED ACTIVE 27263493 CONCAT - RW
sd rootdisk-04 appsvol-01 rootdisk 41948280 27263493 0 c0t0d0 ENA
pl appsvol-02 appsvol ENABLED ACTIVE 27263493 CONCAT - RW
sd rootmirror-04 appsvol-02 rootmirror 41948280 27263493 0 c0t1d0 ENA
v rootvol - ENABLED ACTIVE 16779312 ROUND - root
pl rootvol-01 rootvol ENABLED ACTIVE 16779312 CONCAT - RW
sd rootdisk-01 rootvol-01 rootdisk 0 16779312 0 c0t0d0 ENA
pl rootvol-02 rootvol ENABLED ACTIVE 16779312 CONCAT - RW
sd rootmirror-01 rootvol-02 rootmirror 0 16779312 0 c0t1d0 ENA
v swapvol - ENABLED ACTIVE 8389656 ROUND - swap
pl swapvol-01 swapvol ENABLED ACTIVE 8389656 CONCAT - RW
sd rootdisk-02 swapvol-01 rootdisk 16779312 8389656 0 c0t0d0 ENA
pl swapvol-02 swapvol ENABLED ACTIVE 8389656 CONCAT - RW
sd rootmirror-02 swapvol-02 rootmirror 16779312 8389656 0 c0t1d0 ENA
v var - ENABLED ACTIVE 16779312 ROUND - fsgen
pl var-01 var ENABLED ACTIVE 16779312 CONCAT - RW
sd rootdisk-03 var-01 rootdisk 25168968 16779312 0 c0t0d0 ENA
pl var-02 var ENABLED ACTIVE 16779312 CONCAT - RW
sd rootmirror-03 var-02 rootmirror 25168968 16779312 0 c0t1d0 ENA
Thanks much for any assistance on this.

I would boot off of a cdrom and then simply use the format command to wipe the disks individually.
alan
http://www.sun.com/blueprints/0600/scrub.pdf

Volume manager: SSVM vs. SDS

Hi everyone,
For some odd reason, we configured our 2-node cluster (V2.2) using DiskSuite. We are in the process of adding another node to this cluster. I have heard that DiskSuite cannot be used for a cluster with more than 2 nodes. I called Sun tech support and was told not true. Anyone out there is using SDS on a 3 0r 4 node cluster?
Is there any significant advantage of reconfiguring the cluster with Veritas Volume manager. Does SSVM offer better performance than SDS? We have about 200GB of data that must be converted if switching from SSD to SSVM.
Thanks in advance.
scj

It depends first on what version of Solaris you are running on your nodes. SSVM and CVM are not labeled as supported yet, for Solaris 7. If you have 2.6 then it is your choice. I run SSVM and CVM in an OPS environment, and I love how easy the administration is from the vxva GUI. Now the following is straight out of the Cluster Install Guide,
pg 2-9. If you only run with two disk expansion units, you will need to use Solstice DiskSuite mediators. And then the next bullet says, when using mediators only 2 nodes can act as mediator hosts. I'll let you interpret that depending on your config.
Hope this helps.
Heath
[email protected]

Solaris Volume Manager or Hardware RAID?

Hi - before I build some new Solaris servers I'd like thoughts on the following please. I've previously built our Sun servers using SVM to mirror disks and one of the reasons is when I O/S patch the server I always split the mirrors beforehand and in the event of a failure I can just boot from the untouched mirror - this method has saved my bacon on numerous occasions. However we have just got some T4-1 servers that have hardware RAID and although I like this as it moves away from SVM / software RAID and to hardware RAID I'm now thinking that I will no longer have this "backout plan" in the event of issues with the O/S updates or otherwise however unlikely.
Can anyone please tell me if I have any other options?
Thanks - Julian.

Thanks - just going through the 300 page ZFS admin guide now. I want to ditch SVM as it's clunky and not very friendly whenever we have a disk failure or need to O/S patch as mentioned. One thing I have just read from the ZFS admin guide is that:
"As described in “ZFS Pooled Storage” on page 51, ZFS eliminates the need for a separate volume
manager. ZFS operates on raw devices, so it is possible to create a storage pool comprised of
logical volumes, either software or hardware. This configuration is not recommended, as ZFS
works best when it uses raw physical devices. Using logical volumes might sacrifice
performance, reliability, or both, and should be avoided."
So looks like I need to destroy my hardware RAID as well and just let ZFS manage it all. I'll try that, amend my JET template and kick of an install and see what it looks like.
Thanks again - Julian.

Solaris9 Logical Volume Manager

Dear All,
I want to know more about Logical Volume Manager in Solaris9.
I know that it makes logical volumes from one disk or more?
are these Logical Volumes like a disk can be used for striping datafiles into them if I use RAID 5 controller in Oracle Database?

no replies???????

How to overwrite expired volumes with lower sequence number?

I test expiration policy.
I have a tape library with two volumes.
I defined expiration policy after some time when write windows closed.
lsv -L test_lib
in 1: volume mf_test-000001, barcode e61bcfea534e10281bb000b5df3a54e, 4148676704 kb remaining, expired
in 2: volume mf_test-000002, barcode e6251c26534e10281bb000b5df3a54e, 4148657504 kb remaining, expires 2011/04/28.11:49
How to make OSB overwrite 1st expired volume when writewindow closed on the 2nd volume (with lower sequence number)?
Is it possibly without rotation policy?
If not how to set the policy in my case?
Thanks!

Thanks, Rich!
But I have now following situation:
ob> lsv -a
VOID OOID Seq Volume ID Barcode Family Created Attributes
104 104 1 mf_test-000001 b927cfe85411102ab8e000b5df3a54e mf_test 04/29.10:53 expired
107 107 1 mf_test-000004 b93699ec5411102ab8e000b5df3a54e mf_test 04/29.13:32 closed; expires 04/29.14:32
ob> lsj -l
oracle/11:
Type: database emgc (dbid=1838014073)
Scheduled time: none
State: processed; Oracle job(s) scheduled
Priority: 100
Run on host: (administrative server)
Attempts: 0
oracle/11.1:
Type: datafile backup
Backup piece: 2jmb0c82_1_1
Family: mf_test
Encryption: awaiting job completion
Scheduled time: none
State: pending resource availability
Priority: 100
Run on host: dbsora98
Deferred because: a required device is not available
Attempts: 2
ob> catxcr --tail 100 oracle/11.1
2011/04/29.14:19:49 ______________________________________________________________________
2011/04/29.14:19:49
2011/04/29.14:19:49 Transcript for job oracle/11.1 running on dbsora98
2011/04/29.14:19:49
2011/04/29.14:19:49 ---
2011/04/29.14:19:49 No volume could be found in vl_test (for drive vd_test_2) to use for job oracle/11.1.
2011/04/29.14:19:49 A volume usable for media family mf_test is needed to perform a backup.
2011/04/29.14:21:24 ______________________________________________________________________
2011/04/29.14:21:24
2011/04/29.14:21:24 Transcript for job oracle/11.1 running on dbsora98
2011/04/29.14:21:24
2011/04/29.14:21:24 ---
2011/04/29.14:21:24 No volume could be found in vl_test (for drive vd_test_1) to use for job oracle/11.1.
2011/04/29.14:21:24 A volume usable for media family mf_test is needed to perform a backup.
Edited by: OraSun on 29.04.2011 1:38
I'm sorry, that's ok. other job is holding the 1st volume.

Automounter and volume manager based on DevKit

Hi there,
I've written a simple automounter and volume manager based on DeviceKit:
http://www.ashrentum.net/jmcejuela/prog … 0.2.tar.gz
Some may find it interesting.
Currently it only manages media drives. I hasn't support yet for CDs/DVDs.

jmcejuela wrote:
Hi there,
I've written a simple automounter and volume manager based on DeviceKit:
http://www.ashrentum.net/jmcejuela/prog … 0.2.tar.gz
Some may find it interesting.
Currently it only manages media drives. I hasn't support yet for CDs/DVDs.
Hello, good job!
I was thinking of writing a patch for skvm to use DeviceKit. How does DeviceKit compare to HAL? any obvious deficiencies or whatnot? I've not looked into this at all yet. I'll look into the code to see if I can get any ideas

Volume Manager in Solaris 10

Hi All -
I loaded Solaris 10 OS on two separate disks to be managed by RAID. The idea was that if one disk OS crashes, then RAID will switch me to the other disk and I should be able to boot from that without major loss in operational time of the node. For such a scenario, do I need to have Veritas Volume Manager or does Solaris 10 come with an default volume manager which will do this for me.
Your help is greatly appreciated.
Regards

RTFM :-)
http://docs.sun.com/app/docs/doc/816-4520
alan

How do you change volume permissions with Solaris Volume Manager?

(Previously posted in "Talk to the Sysop" - no replies)
I'm trying to set up Solaris 9 to run Oracle on raw partitions. I have my design nailed down and I have built all the raw partitions I need as soft partitions on top of RAID 1 volumes. All this is built using Solaris Volume Manager (SVM).
However, all the partitions are still owned by root. Before I can create my Oracle database, I need to change the owner of the Oracle partitions to oracle:oinstall. The only reference I found telling me how to do this was in a Sun Blueprint and it essentially said "You can't change volume permissions directly or permanently using SVM and chown will only remain effective until the next reboot. To make the changes permanent, you must modify /etc/minor_perm". Unfortunately, I can't find an example of how to do this anywhere and the online man pages are not particularly helpful (at least not to me).
I'd appreciate a quick pointer, either to a good online resource or, even better, a simple example. For background, the volumes Oracle needs to own are:
/dev/md/rdsk/d101-109
/dev/md/rdsk/d201-203
/dev/md/rdsk/d301-303
/dev/md/rdsk/d401-403
/dev/md/rdsk/d501-505
I provide this information because I'd like to assign some, but not all, of the devices under /dev/md/rdsk to the oracle user and I was hoping some smart person out there could illustrate an approach using simle regular expressions, at which I'm horribly poor.
Thanks in advance,
Adrian

Ron, I feel your pain. I just came from an HTC also and a lot of stuff with this iPhone is bugging the crap out of me. Who makes a phone where you can't adjust the ringer and alert volumes independently? Instead, I have to adjust the alert volume when it is active. C'mon guys. Get with the program. You won a bunch of Android users over with the 4S, but you're going to chase us all back when we're done with our contract. Frustrating.

Volume Manager - Striping - sequence writing performance

Similar Messages

Maybe you are looking for