unixadmin.free.fr Handy Unix Plumbing Tips and Tricks

9oct/20Off

Hints and Tips for Migrating Workload to IBM POWER9

Salut les Z'admins Unix :D

Un petit retour d'expérience sur ce qu'il faut prévoir avant de migrer des partitions AIX sur POWER9 (ex E980).

A lire et à appliquer absolument :
Hints and Tips for Migrating Workload to IBM POWER9 Processor-Based Systems
- appliquer les APARS listés pour AIX et Virtual I/O Server

Les correctifs "P9 VPM fold threshold" sont importants. Sur Power9 pour AIX 6.1 TL09 SP12, si vous n'avez pas le kernel bos.mp64 6.1.9.403, vous pouvez appliquer les valeurs schedo recommandées par ce script : http://gibsonnet.net/blog/cgaix/resource/p9_vpm_perf_script.txt

La modification des paramètres vpm_fold_threshold et vpm_xvcpus apportent de nette amélioration pour les "workloads DB Oracle" AIX 6.1 SMT4 sur Power9 E980 sans le kernel 6.1.9.403.

schedo -p -o vpm_fold_threshold=36
schedo -p -o vpm_xvcpus=2

Ensuite il y a aussi ces lectures très intéressantes qui permettent de mieux comprendre l'optimisation des workloads AIX sur P9, surtout lorsque l'on mutualise plein de LPAR aux tempéraments différents.

Le folding cpu
une très claire explication sur le VPM FOLDING
Source: IBM POWER Virtualization Best Practices Guide

Le Placement des ressources utilisées par la LPAR
A consulter avec lssrad (AIX)

root@lpar074 # lssrad -av

REF1 is the second level domain (Drawer or DCM).
SRAD refers to processor chips.
The “lssrad” results for REF1 and SRAD are logical values and cannot be used to determine actual physical chip/drawer/DCM information.

Dynamic Platform Optimizer
L'efficacité de DPO n'est plus à prouver surtout quand le score de la LPAR et le lssrad ne sont pas au top, voir lsmemopt et optmem sur la HMC.
Source: Performance Optimization and Tuning Techniques for IBM Power Systems Processors Including IBM POWER8

Affinity groups
This affinity group feature can be used in multiple situations:
LPARs that are dependent or related, such as server and client, and application server and database server, can be grouped so they are in the same book.
Pour le Graal. A ce niveau d'optimisation on peut entendre les chevaliers de la Table ronde.
Source: www.capacityreports.net: Optimising Power with Affinity Groups

Comment vérifier le placement des LPAR dans les CEC drawer :

hscroot@hmc:~> startdump -m E980 -t resource -r "hvlpconfigdata -base -procs -memory -domain -affinity"
hscroot@hmc:~> cat /dump/RSCDUMP.XXXXXX.0X000000.20201011160712

Récupérer les ID des LPARs

hscroot@hmc:~> lssyscfg -m E980 -r lpar -F lpar_id name state os_version
150 lpar021 Running "AIX 7.2 7200-03-03-1914"
151 lpar011 Running "AIX 7.2 7200-03-03-1914"
24 lpar023 Running "AIX 6.1 6100-09-12-1846"
25 lpar013 Running "AIX 6.1 6100-09-12-1846"
...
4 vio4 Running "VIOS 3.1.1.25"
3 vio3 Running "VIOS 3.1.1.25"
2 vio2 Running "VIOS 3.1.1.25"
1 vio1 Running "VIOS 3.1.1.25"
...

Extraire le tableau suivants du dump RSCDUMP.XXXXXX.0X000000.20201011160712

|-----------|-----------------------|---------------|------|---------------|---------------|-------|
|   Domain  |     Procs       Units |     Memory    |      |   Proc Units  |     Memory    | Ratio |
| SEC | PRI | Total | Free  | Free  | Total |  Free |  LP  |  Tgt  |  Aloc |  Tgt  |  Aloc |       |
|-----|-----|-------|-------|-------|-------|-------|------|-------|-------|-------|-------|-------|

SEC = CEC Drawer ( de 0 à 3 pour un E980 Power9 4 x CEC )
PRI = P9 processor chips
LP = LPAR ID

Source: Source: www.capacityreports.net

La désactivation de Spectre/Meltdown pour favoriser la performance à la sécurité.
Source: https://www.ibm.com/support/pages/how-disableenable-switch-spectremeltdown-patches-power8-systems-asmi-menu

Voila, en appliquant ces bonnes pratiques avec discernement vous devriez pouvoir redonner le sourire à vos chers utilisateurs.
Merci à Youssef, Sami, Laurent et Aleksandar pour leurs conseils :D

Portez vous bien,
Samir

Remplis sous: AIX, IBM SYSTEM Commentaires
10fév/16

How to check memory and core activated via CUoD Activation Code

Go to IBM capacity on demand, enter type and serial number and check POD and MOD lines.

Ex 1: model 9117 type MMD
POD 53C1340827291F44AAF4000000040041E4 09/27/2015
AAF4 = CCIN = 4.228 GHz core
04 = 4 core activated

MOD 2A2A7F64BEEEC606821200000032004187
8212 = Feature code = Activation of 1 GB
32 = 32 GB activated

Ex 2: model 8233
POD 80FF07034C0917FA771400000016004166 09/17/2010
7714 = Feature code = 3.0 GHz core
16 = 16 core activated

Source :
Thank's to Mr Delmas
for CCIN reference check IBM Knowledge Center
for Feature code reference check IBM sales manual

7jan/15

How to check for memory over-commitment in AME

Question

In LPARs that utilize the Power 7 (and later) feature of Active Memory Expansion (AME), assessing memory resources is a more complex task compared with dedicated memory systems. How can the memory in such a system be evaluated?

Answer

Introduction

Active Memory Expansion (AME) allows for the compression of memory pages to increase the system's effective virtual address space. At high usage, unused computational memory is moved to the compressed pool instead of being paged-out to paging space. This is typically employed in environments with excess CPU resources and are somewhat constrained on physical memory. Active Memory Expansion is feature that has been introduced in POWER7/POWER7+ systems with a minimum level of AIX 6.1 TL4 SP2.

AME Scenarios

After planning and configuring the system with the amepat tool, there are some scenarios that might require a change of AME configuration:

  1. Virtual memory exceeds Target Memory Expansion Size

  2. When this scenario is present, the system is over-committed and will start paging out to disk. From a configuration stand-point, rerun the amepat tool to either increase the Expansion Factor or to increase the size of physical memory.

  3. Virtual memory exceeds assigned physical memory and less than Target Memory Expansion Size (with no deficit)

  4. This is the ideal scenario when using AME as the compressed pool is able to satisfy the memory demands of the LPAR.

  5. Virtual memory exceeds assigned physical memory and less than Target Memory Expansion Size (with deficit)

  6. When the system is unable to compress memory pages to meet the Target Memory Expansion Size, there will be a deficit and pages that exceed the allocated memory are moved to paging space. Not all memory pages are subject to compression (pinned pages or client pages) and therefore, a deficit is present. Rerun the amepat tool to either decrease the Expansion Factor or to increase the size of physical memory.

  7. Virtual memory is below assigned physical memory

  8. While there isn't a problem with over-commitment with this setup, it is not benefiting from AME. Rerun the amepat tool to decrease the allocated physical memory and evaluate the current Expansion Factor.

Tools to use with a live example
The following tools on AIX can be used to determine the current status of an AME-enabled LPAR (with a live example from the IBM Redbook IBM PowerVM Virtualization Managing and Monitoring):

# amepat

  1. Comparing the Virtual Memory Size (MB) to the Target Expanded Memory Size, we find that the system is not over-committed logically.
  2. Due to the Deficit Memory Size (MB), the system will start utilizing paging space due to the inability to compress more memory.

# vmstat -c

Comparing the avm value (in 4k pages) to the tmem value (MB) will tell us if the system is logically over-committed.

  1. Observing the dxm will show us the deficit in 4k pages.

# svmon -O summary=AME

Comparing the virtual column to the size column shows no issue with logical memory over-commitment.

  1. The dxm column shows the deficit in 4k pages

For more information regarding AME, please refer to the IBM Redbook IBM PowerVM Virtualization Managing and Monitoring (sg247590):
http://www.redbooks.ibm.com/abstracts/sg247590.html

Source: IBM Technote

Taggé comme: Aucun commentaire
21oct/13Off

Debian Linux sur IBM POWER … easy :)

L'installation de Debian Linux 7.2 sur IBM POWER est déconcertant de facilité. Ci-dessous un exemple d'installation via une image ISO "Debian netinstaller" debian-7.2.0-powerpc-netinst sur Virtual I/O Server shared storage pool

Création du disque système debian1 d'une taille de 10Go et mapping vers le vhost8 sous le nom debian_sda

$ mkbdsp -clustername CL570 -sp SPA 10G -bd debian1 -vadapter vhost8 -tn debian_sda
Lu Name:debian1
Lu Udid:72cf212efed9a724585b307a92a83ce0

Assigning file "debian1" as a backing device.
VTD:debian_sda

$ lsmap -vadapter vhost8
SVSA            Physloc                                      Client Partition ID
--------------- -------------------------------------------- ------------------
vhost8          U9117.570.658502E-V2-C90                     0x00000000

VTD                   debian_sda
Status                Available
LUN                   0x8100000000000000
Backing device        debian1.72cf212efed9a724585b307a92a83ce0
Physloc
Mirrored              N/A

Création d'un repository pour stocker l'image ISO "debian-7.2.0-powerpc-netinst.iso"

$ mkrep -sp rootvg -size 10G
Virtual Media Repository Created
Repository created within "VMLibrary" logical volume

$ mkvopt -name debianISO -file /home/padmin/ISO/debian-7.2.0-powerpc-netinst.iso –ro

df -g /var/vio/VMLibrary
Filesystem    GB blocks      Free %Used    Iused %Iused Mounted on
/dev/VMLibrary     10.00      9.71    3%        5     1% /var/vio/VMLibrary

Création d'un DVD virtuel et chargement de l'image ISO

$ mkvdev -fbo -vadapter vhost8
vtopt0 Available

$ loadopt -vtd vtopt0 -disk debianISO

$ lsmap -vadapter vhost8
SVSA            Physloc                                      Client Partition ID
--------------- -------------------------------------------- ------------------
vhost8          U9117.570.658502E-V2-C90                     0x00000000

VTD                   debian_sda
Status                Available
LUN                   0x8100000000000000
Backing device        debian1.72cf212efed9a724585b307a92a83ce0
Physloc
Mirrored              N/A

VTD                   vtopt0
Status                Available
LUN                   0x8200000000000000
Backing device        /var/vio/VMLibrary/debianISO
Physloc
Mirrored              N/A

Boot su CD via le menu SMS

IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM
IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM
IBM IBM IBM IBM IBM IBM                             IBM IBM IBM IBM IBM IBM
IBM IBM IBM IBM IBM IBM     STARTING SOFTWARE       IBM IBM IBM IBM IBM IBM
IBM IBM IBM IBM IBM IBM        PLEASE WAIT...       IBM IBM IBM IBM IBM IBM
IBM IBM IBM IBM IBM IBM                             IBM IBM IBM IBM IBM IBM
IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM
IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM IBM


Elapsed time since release of system processors: 15905 mins 2 secs

Config file read, 1337 bytes
Welcome to Debian GNU/Linux wheezy!

This is a Debian installation CDROM,
built on 20131012-15:01.


Enter one of the following options to begin:

install     64-bit processor (G5 or POWER3/4/5/6/7)
install32   32-bit processor (G4 or earlier)

.....
Welcome to yaboot version 1.3.16
Enter "help" to get some basic usage information
boot:install

Next ... Next ... Next ... Reboot (à ne pas faire sous Windows)

Bon, visiblement la bootlist est toujours sur le CD.
Donc, arrêt au menu SMS et paramétrer la bootlist manuellement vers le disque virtuel.

root@deb570:~# uname -a
Linux deb570 3.2.0-4-powerpc64 #1 SMP Debian 3.2.51-1 ppc64 GNU/Linux

Longue vie à tous les développeurs Debian PowerPC ;)

Taggé comme: Commentaires
15juil/13

IBM Power Facts and Features

IBM Power Systems, IBM PureFlex and Power Blades
June 2013

IBM Power Facts and Features

Taggé comme: Aucun commentaire
26avr/11

POWER7 Systems and AIX supported levels

Source IBM Technote

Synthetic view of POWER7 Systems and AIX supported levels (updated with Services Packs' Fall 2010 availability)

Download
POWER7_Systems_AIX_levels_2010-11.xls

10mar/11

upgrade firmware P5 standalone avec CD de diag

Mise à jour Microcode d'un serveur P5 standalone SF230 ou SF235 sans système AIX et à intégrer sur HMC V7.3.5

1. Se connecter au menu ASMI via port série (laptop + hyperterminal)
Login : admin
password : admin

System Configuration
Firmware Update Policy: Operating system

Power/Restart Control
Power On/Off System
Firmware boot side for the next boot: Temporary

2. Sur serveur AIX de test
Décompresser le Rpm :

 # rpm -Uhv --ignoreos 01SF240_382_382 .rpm

Remarque : l’image sera toujours décompressée dans /tmp/fwupdate.

Renommer l'image en .img

# mv 01SF240_382_382  01SF240_382_382.img

Noter le résultat du checksum

# sum /tmp/fwupdate/01SF240_382_382.img

3. Récupérer l'image 01SF240_382_382.img via FTP en mode binaire

4. Graver 01SF240_382_382.img sur un CDrom avec le fichier 01SF240_382_382.xml

Vérifier le checksum après la gravure sur un AIX de test

# mount -v cdrfs -oro /dev/cd0 /mnt

Comparer le checksum

# sum /mnt/01SF240_320_201.img
# umount /mnt

5. Supprimmer le rpm du serveur de test AIX

# rpm -qa | grep 382
# rpm -e 01SF240_382_382

6. Upgrade firmware via CD diag 53060
ftp://ftp.software.ibm.com/software/server/diags/CD53060.iso'

Mise à jour microcode avec le CD de diag
Démarrer via le menu SMS sur le CD de diag
- Task Selection
- Microcode Tasks
- Update and Manage Flush

... Insérer le CD contenant le microcode une fois demandé.

7. Avant de rattacher le P5 à la HMC positionner le firmware update policy

Se connecter au menu ASMI
Login : admin
password : admin

System Configuration
Firmware Update Policy: HMC

Power/Restart Control
Power On/Off System
Firmware boot side for the next boot: Temporary

8. Ajouter le P5 à la HMC V7 via les menus de la HMC

9mar/11

COMPARATIF P5 et P6 SERVEUR IBM / BULL

Model IBM
Reference Proc. Freq Reference Bull
8203-E4A P520 P6 4.2Ghz PL160R (1proc)
8203-E4A P520 P6 4.2Ghz PL160T (1proc)
8203-E4A P520 P6 4.2Ghz PL260R (2proc)
8203-E4A P520 P6 4.2Ghz PL260T (2proc)
8203-E4A P520 P6 4.2Ghz PL460R (4proc)
8203-E4A P520 P6 4.2Ghz PL460T (4proc)
8204-E8A P550 P6 3.5/4.2Ghz PL860R (2,4,6,8 proc)
8204-E8A P550 P6 3.5/4.2Ghz PL860T (2,4,6,8 proc)
9117-MMA P570 P6 3.5/4.2/4.7Ghz PL1660R (216proc)
9119-FHA P595 P6 4.2/5.0 Ghz PL3260R (32 proc)
9119-FHA P595 P6 4.2/5.0 Ghz PL6460R (64 proc)
9110-51A P510Q P5+ 1.65Ghz PL450R-VL+
9110-51A P510Q P5+ 1.9/2.1Ghz PL250R-L+
9110-510 P510 P5 1.5/1.65 GHz PL250R-L
9111-520 P520 P5 1.5/1.65 GHz PL250T, PL250R
9131-52A P520 P5+ 1.65/1.9/2.1 GHz PL250T+, PL250R+
9131-52A P520Q P5+ 1.5/1.65Ghz PL450T-L+, PL450R-L+
9113-550 P550 P5 1.5/1.65 GHz PL450T, PL450R
9133-55A P550 P5+ 1.65/1.9/2.1 GHz PL450T+, PL450R+
9133-55A P550Q P5+ 1.5/1.65 GHz PL850T-L+, PL450R-L+
9115-505 P505 P5 1.5/1.65 GHz PL250R-VL
9115-505 P505 P5+ 1.9/2.1Ghz PL250R-VL+
9115-505 P505 P5+ 1.65Ghz
9116-561 P560 P5+ 1.5/1.8 GHz PL1650R-L+
9117-570 P570 P5 1.5/1.65/1.9 GHz PL850R, PL1650R
9117-570 P570 P5+ 1.9/2.2 GHz PL850R+, PL1650R+
9119-590 P590 P5 1.65Ghz PL3250R
9119-595 P595 P5 1.65/1.9Ghz PL6450R
9119-590 P590 P5+ 2.1Ghz PL3250R+
9119-595 P595 P5+ 2.1/2.3Ghz PL6450R+
2jan/11

Resetting the managed system to a nonpartitioned configuration

You can use the Hardware Management Console (HMC) and the Advanced System Management Interface (ASMI) to erase all of your logical partitions and reset the managed system to a nonpartitioned configuration. When you reset the managed system, all of the physical hardware resources are assigned to a single logical partition. This allows you to use the managed system as if it were a single, nonpartitioned server.
Attention: By resetting a partitioned managed system to a nonpartitioned configuration, you will lose all of your logical partition configuration data. However, resetting the managed system does not erase the operating systems and data from disk units on that managed system.

Before you reset the managed system, ensure that the hardware placement in the managed system supports a nonpartitioned configuration. If the hardware placement in the managed system does not support a nonpartitioned configuration, you must move the hardware so that the hardware placement supports a nonpartitioned configuration. For more information about how to place the hardware in your managed system to support a nonpartitioned configuration, contact your marketing representative or business partner.

Also, if you plan to use an operating system that is already installed on one of the logical partitions on the managed system (instead of reinstalling the operating system after you reset the managed system), consider how the console used by that operating system will change when you reset the managed system. If the operating system that you want to use is AIX®, log into AIX and enable the login prompt for the virtual serial port vty0 using either the System Management Interface Tool (SMIT) or the chdev command. You can then reset the managed system, use a physical serial console to log into AIX, and use SMIT or the chcons command to change the console device to the console device you want to use.

You must have an ASMI login profile with an administrator authority level.

Parts of this procedure must be performed at your HMC (not connected remotely). Ensure that you have physical access to the HMC before you begin.

To reset a managed system with logical partitions to a nonpartitioned configuration using the HMC, follow these steps:

Shut down all logical partitions on your managed system using operating system procedures. For more information about shutting down logical partitions using operating system procedures, see the following information:
For logical partitions running AIX, see Shutting down AIX logical partitions.
For logical partitions running IBM® i, see Shutting down IBM i logical partitions.
For logical partitions running Linux®, see Shutting down Linux logical partitions.
For logical partitions running Virtual I/O Server, see Shutting down Virtual I/O Server logical partitions using the HMC.

If the managed system powered off automatically when you shut down the last logical partition, power on the managed system to the Standby state. Complete the following:

- In the navigation pane of your HMC, open Systems Management and click Servers.
- In the work pane, select the managed system, click the Tasks button, and click Operations > Power On.
- Select the power-on mode of Partition Standby and click OK.
- Wait until the work pane displays a Standby state for the managed system.
- Initialize the profile data on the HMC. Complete the following:
- In the work pane, select the managed system, click the Tasks button, and click Configuration > Manage Partition Data > Initialize.
- Click Yes to confirm.
- Clear the logical partition configuration data on the managed system. Complete the following at your HMC (not connected remotely):
- In the navigation pane, click HMC Management.
- In the work pane, click Open Restricted Shell Terminal.
- Type the command:

lpcfgop -m managed_system_name -o clear

- where managed_system_name is the name of the managed system as it is displayed in the work pane.
- Enter 1 to confirm. This step takes several seconds to complete.

Optional: If you no longer intend to manage the system using the HMC, remove the connection between the HMC and the managed system. To remove the connection between the HMC and the managed system, complete the following:
- In the work pane, select the managed system, click the Tasks button, and click Connections > Reset or Remove Connection.
- Select Remove connection and click OK.
- Access the Advanced System Management Interface (ASMI) using a Web browser on a PC. If you do not already have a PC that is set up to access the ASMI on the managed system, you need to set up the PC at this point. For instructions, see Accessing the ASMI using a Web browser.
- On the ASMI Welcome pane, log in using the admin user ID (enter admin into User ID, enter the admin password into Password, and click Log In).
- In the navigation pane, expand Power/Restart Control and click Power On/Off System.
- Set Boot to server firmware to Running.
- Click Save settings and power off.
- Click Power On/Off System periodically to refresh the window. Repeat this step until Current system power state: Off is displayed in the navigation pane.
- Click Save settings and power on.
- Wait for the managed system to restart. It can take several minutes for the managed system and operating system to restart completely.

2fév/10

Recuperer un RAID5 SCSI sur carte scraid IBM

Suite coupure électrique sur baie 2104 , perte du Groupe de volume .... redémarrage de la baie et récupération du Raid5.

# hdisk2 define 20-58-00-0,0 SCSI Disk Array RAID 5

# smit pdam

List PCI SCSI Disk Arrays ou scraidmgr -l hdisk2
hdisk2  Available Raid  5 10-70-00-0,0 8606 MB Status DEAD
      hdisk2  2C Channel 2 ID C  FAILED DRIVE - 4304Meg
      hdisk2  29 Channel 2 ID 9  FAILED DRIVE - 4304Meg
      hdisk2  2A Channel 2 ID A  FAILED DRIVE - 4304Meg

Option "revive a failed drive" sur tous les disques physiques

Revive a FAILED Drive in a PCI SCSI Disk Array
    Select a Logical Drive    
        hdisk2
        Revive a FAILED Drive  
             hdisk2  2C Channel 2 ID C  FAILED DRIVE - 4304Meg
             hdisk2  29 Channel 2 ID 9  FAILED DRIVE - 4304Meg
             hdisk2  2A Channel 2 ID A  FAILED DRIVE - 4304Meg

Impossible d'effectuer "revive a failed drive" sur le dernier disque

hdisk2  Available Raid  5 10-70-00-0,0 8606 MB Status DEGRADED
      hdisk2  2C Channel 2 ID C  ONLINE - 4304Meg
      hdisk2  29 Channel 2 ID 9  ONLINE - 4304Meg
      hdisk2  2A Channel 2 ID A  FAILED DRIVE - 4304Meg
 Reconstruct a PCI SCSI Disk Array
    Select a Logical Drive    
        hdisk2
       
    Reconstruct a PCI SCSI Disk Array
         2A Channel 2 ID A  FAILED DRIVE - 4304Meg
           
    OK 100 percent complete
List PCI SCSI Disk Arrays
hdisk2  Available Raid  5 10-70-00-0,0 8606 MB Status OPTIMAL
      hdisk2  2C Channel 2 ID C  ONLINE - 4304Meg
      hdisk2  29 Channel 2 ID 9  ONLINE - 4304Meg
      hdisk2  2A Channel 2 ID A  ONLINE - 4304Meg
# cfgmgr

# hdisk2 available 20-58-00-0,0    SCSI Disk Array RAID 5

Si la carte raid scsi à été remplacée
1. Select the Recovery Options option
3. Select Resolve PCI SCSI RAID Adapter Configuration option.
4. Select Accept Configuration on Drives option.
5. Select the adapter that was replaced.

Si les drives reste failed
1. Select the Recovery Options option of the PDAM.
2. Select Resolve PCI SCSI RAID Adapter Configuration.
3. Select Display/Accept Configuration Changes.
4. Select the adapter reporting the conflict.
5. Change the Display only option to NO.