Blog du DS45: juillet 2009

vendredi 31 juillet 2009

HOWTO SCO P2V in the real life

Have you ever try to P2V a SCO Open Server 5.X.X machine on vSphere ? Oh, .... sorry.... What is SCO Open Server ? It was the state of art of Unix on x86 15 years ago. I used to install few hundred at that time.

Many efficient applications, so efficient they already run, were installed at that time, but today the hardware is really end of life, people who installed it out of work and applications a must have .....
Reading VMware documentation, it seems to be quite easy but it's not : You can P2V, but not reconfigure the HD image and finally have a kernel panic when you boot your new virtual machine because Unix is very near from the Hardware. For example an initial hardware Compaq Proliant RAID controller could not run as an emulated one, and a Buslogic BTLD - Boot Time Loadable Driver in the SCO World - does not run better in vSphere as you could find in some blogs because it's the IDE that is recognize in vSphere. So let's try another way ....

1st, download an OS image from Sco's ftp site : ftp://ftp.sco.com/pub/openserver5/507/iso/OpenServer-5.0.7-July2005/OpenServer-5.0.7Hw-10Jun05_1800.iso

and Nics update with Intel Pro 1000 support

ftp://ftp.sco.com/pub/openserver5/507/drivers/eeG_5.0.7g/VOL.000.000, save it as VOL.000.000.flp

Create a VM as other OS : Define 1 CPU, 256Mb of ram, choose BusLogic controller, X Gb of Virtual HD as needed, choose Intel Pro 1000 as Nic, Hardware v7 even if you find SCO support with v7 : there is no VMware tools compliant in vSphere ....

Boot from .ISO and install it. 

On the physical machine, you need several informations

- hostname, ip adress, netmask using 'netconfig' at prompt, navigate from up to center window using tab key.
- get default gateway with a cat /etc/default/tcp GATEWAY=
- find the HD driver with a hwconfig at prompt and get th TYPE= where you will find the actual hardware HD driver. For example, 'cha' is for Compaq Smart Array controller.

If you got enough space, create a backup file of your entire disk inside the file system with :
cd / ; find ./ -print cpio -ovfB > /backupfile.cpio &
If you filesystem is full, mount a network drive via NFS and backup onto :
mount -f NFS X.X.X.X:/share /local_mount_directory
cd / ; find ./ -print cpio -ovfB > /local_mount_directory/backupfile.cpio &

If NFS client package is not installed, put the SCO Installation CD in drive and install it with the 'custom' command.

Be careful, you could not create a backup file larger more than 2GB.... it's the old world. If so, make .cpio file for each directory you need.

Boot your fresh Openserver image.

Press CTRL-D to boot in multi user level, login with root account (give the password created during installation) Get with ftp all backup .cpio files. Now, install them with

cpio –ivdmlu < /backupfile.cpio &

You will probably have some time to spend drinking coffee :)

When it’s finished, don’t reboot !!

Mount from vCenter VOL.000.000.flp as floppy in your Openserver image. Use ‘custom’ to install from floppy image, INTEL Pro 1000 driver.

When you will exit 'custom' with tab key/file/exit, an Unix kernel will be relinked.

Now it's time to configure networking

launch 'netconfig' command and select hardware/Add New lan Adapter

Choose TCP/IP, forget IPX/SPX is an old Novell Network Protocol

Configure IP adress as usual, exit and edit /etc/default/tcp to indicate GATEWAY= X.X.X.X as default gateway.

Update : It's possible to use "Flexible Ethernet" as Ethernet option in your VM description instead of PRO1000. Your SCO will see this NIC as AMD PCNet-PCI adapter without installing anything. It seems to works faster ....

Declare actual disk driver with 'mkdev hd 0 SCSI-0 0 blc 0

FYI :

- blc is the SCO's BusLogic driver which support vSphere's virtual one
- mkdev hd [TARGET] [HA/Controller number] [LUN] [Driver] [Bus Number] will cause another kernel relink.

Edit /etc/default/boot

add disable=[old driver] Sdsk=(0,0,0,0) at the end of defbootstr :

ie : DEFBOOTSTR=hd(40)unix swap=hd(41) dump=hd(41) root=hd(42) disable=cha Sdsk=blc(0,0,0,0)

Your old controller driver will be disable and dynamically replaced by the supported BusLogic one.

Reboot with : sync;sync;sync;reboot or sync;sync;sync;haltsys
If you forgot the edit part of /etc/default/boot, don't worry, add 'defbootstr disable={ctl driver] Sdsk=blc(0,0,0,0)' after boot: prompt ( be careful, you are in QWERTY world ...)

You may delete old disk driver from the kernel with editing 2 files :
/etc/conf/cf.d/mscsi, enter "blc Sdsk 0 0 0 0" and remove lines with the old driver

In /etc/conf/sdevice.d, edit file with old driver name such as "cha" and remplace "Y" by "N" - :1,$s/Y/N/g in vi .... which means from line 1 to end of file, substitue Y by N in all file :)

In /etc/conf.d/sdevice.d/wd, verify the 2 lines are validated with "Y"

Relink the kernel /etc/conf/cf.d/link_unix and reboot.

You may delete old disk driver from the kernel with editing 2 files :
/etc/conf/cf.d/mscsi, enter "blc Sdsk 0 0 0 0" and remove lines with the old driver

In /etc/conf/sdevice.d, edit file with old driver name such as "cha" and remplace "Y" by "N" - :1,$s/Y/N/g in vi .... which means from line 1 to end of file, substitue Y by N in all file :)

In /etc/conf/sdevice.d/blc edit the file and remplace "N" by "Y" - :1,$s/Y/N/g in vi

Finally relink the kernel /etc/conf/cf.d/link_unix and reboot.

update : If you installed OpenServer on ide (0:0) drive, it's possible to use 'wd' driver instead of BusLogic 'blc' one without BTLD driver

I hope this post was useful, and now enjoy !!

jeudi 30 juillet 2009

Cisco IE3000, un switch 2960 ducit pour les environnements industriels

Petits test menés actuellement par DS sur ce petit matériel bien prometteur.

Les + :
- Modulaire et extensible
- supporte une alimentation 48v pour une consommation de 15w
- Les ports 2 x 2 ports giga cuivre et SFP
- Respecte les normes industrielles, est conforme IP 55, supporte des températures jusqu'à 75°C
- Inclus deux contacts secs
- Se fixe sur les rails DIN (rails des tableaux électriques)
- Supporte Cisco Network Access Control (NAC)
- Contient un emplacement pour une carte flash avec la conf, que l'on peut facilement retirer et remettre dans un autre commutateur de même modèle. C'est pratique pour les équipes de maintenance

Les - :
- ne supporte pas (encore) le protocole MODBUS
- Pas de support POE toujours utile sur au moins un ou deux ports 10/100.
- l'alim dans un boitier séparé

La marrade de la semaine : Google Apps sources d'économies

http://google.fr/appscalculator

- Où sont les pénalités en cas d'indisponibilité ?
- Quid des coûts télécom ?
- Où sont mes données ?
- Le prix des serveurs intègre-t-il la virtualisation, sinon ils sont bien chers .....
- Quelles sont les bases (française) utilisées pour calculer le coût notamment des employés ?

Avec OpenOffice, Evolution (clone de Outlook), Sendmail, Linux et Xen, tout de suite l'écart financier devient beaucoup moins important. Et cerise sur le gâteau, pour bien se rappeler à nous et plutôt que d'envoyer un mail, on préfère envoyer un bon vieux courrier physique : c'est vraiment pas green ça ! ... bref, du gros enfumage marketing.

mercredi 29 juillet 2009

A méditer : piloter, c'est pouvoir prévoir

Did you know Pillar Axiom recently delivered the industry’s most consistent performance results for MS Exchange using the ESRP benchmark? It’s a fact: Three separate tests (12,750, 17,000 and 34,000 users in a single AX600) yielded the same result – an average of 2,125 Exchange mailboxes per FC Brick (ESRP Storage v2.1). The consistent performance of the Axiom demonstrates it scales in performance linearly as capacity is added, providing the predictability needed to properly deploy a successful Exchange solution with no surprises. Boring is better. Click here < http://blog.pillardata.com/pillar_data_blog/2009/07/exciting-or-boring.html > to learn more.

lundi 27 juillet 2009

AH1N1 : le client léger déclaré d'utilité public

Les précautions prises par nos gouvernants ni auront rien fait : nous seront tous atteints par le virus de la grippe porcine techiquement appelé AH1N1. Le pic serait attendu à l'automne en France et il faudra se plier à des mesures strictes pour endiguer la contamination : éviter tout contact; éviter de voyager, se laver fréquemment les mains, voir même se barricader chez soi ... comment faire pour continuer à faire vivre son entreprise à moins d'être au chômage technique.

Il y a quelques semaines, avons été saisis par les autorités nous demandant la mise en place d'un plan de pandémie de niveau 6. Pas facile à appliquer pour une entreprise de service public. D'un point de vue informatique, j'ai suggéré de mettre à contribution notre récente plateforme VMware View 3 / VPN pour permettre à l'ensemble des employés sédentaires de pouvoir travailler à distance. Il suffit de compléter l'existant en ajoutant quelques licenses et serveurs ad-hoc, d'augmenter le débit des lignes d'accès à Internet, le nombre d'accès simultanés VPN, de déployer View et Skype chez nos employés en autoconfigurant la connexion sur leur PC familial depuis la messagerie et les packages APP-V. App-V va bien nous aider sur ce coup-là !! Le coût est finalement marginal comparé à celui d'un arrêt complet de l'activité. Autre aide précieuse, l'installation prochaine d'une nouvelle baie PILLAR DATA full SSD de 1,2To - 2 Go/s de débit et pas un truc de laboratoire pour geek bleu et blanc- permettra de générer quasi instantanément les clones liés. Ainsi, tout un chacun pourra continuer à travailler de manière transparente sur son environnement de bureau habituel et le provisionning se fera réellement à la demande. Bien entendu, ce qui fonctionne dans une entreprise public, est déclinable dans une entreprise privée, mais vous n'êtes pas obligé de me croire.

dimanche 26 juillet 2009

French Transportation Company Consumes 6X Less Energy, Increases Oracle Performance by 10X with Pillar Data Systems

Able to Guarantee Stronger SLAs with Axiom's Application-Aware Storage
San Jose, Calif. – July 15, 2009 – SETAO replaced its NetApp NAS system with the Pillar Axiom™ to achieve better storage efficiency and application performance. Société d'Exploitation du Tramway et du réseau de bus de l'Agglomération d'Orléans (SETAO) manages the urban transportation network of the city of Orléans, France which transports about 100,000 passengers per day. The company trusted Pillar to support its strict reliability and scalability requirements.
“Storage efficiency was a very important factor when we chose the Axiom,” said Olivier Parcollet, IT Director, SETAO. “We would have needed 4,000 Watts with other systems – all with the same capacity. To run the Axiom, we only need 650 Watts of electricity. That’s a very impressive ratio, and one that has already delivered significant cost savings.”
SETAO manages data from trams and buses, vehicle radios, billing systems, electrical systems, traffic lights, and video feeds from surveillance cameras throughout Orléans. In addition to gathering and maintaining this information, the company makes real-time traffic information available via mobile devices and makes surveillance data available for law enforcement. To manage this massive amount of data, the company maintains 15 Oracle databases and other applications on 70 virtual machines.
With Pillar’s Application-Aware capability built on its patented QoS technology, SETAO was able to classify the data from the highest importance to the lowest, effectively allocating resources for maximum storage utilization. “By ensuring certain performance levels, we are able to more accurately warranty a service level agreement to each end user,” said Parcollet.

With the former NetApp storage system, SETAO experienced performance issues due to resource contention, which caused concern about scalability. New drives could not be added. “We would have to stop our existing solution and actually exchange it with the manufacturer to get a system with more capacity,” said Parcollet.

After running tests on NetApp, EMC, and Pillar systems, Parcollet found that an Oracle request ran seven to 10 times faster on the Axiom. And while some of the systems experienced a drastic reduction in performance and reliability when stressed, the Axiom remained stable. The reliability of SETAO’s new system was proven when, just weeks after the deployment, there was an interruption of electricity. The Pillar Axiom remained stable with no downtime.

“When companies like SETAO have hundreds of thousands of people depending on them just to get through their day, they can’t afford any downtime, no matter how short,” said Bob Maness, Vice President, Worldwide Marketing and Channel Sales, Pillar Data Systems. “That’s when they realize they need a storage partner that practices the same philosophy. Combined with its competitive IOPS per terabyte and lower energy requirements, the Axiom improves application performance simply by using storage more efficiently.”

Recently SETAO embarked on a new project to store and manage video from the municipal surveillance cameras—initially 300 video cameras but soon to be expanded to 500. With this huge amount of data and a requirement to retain it for at least 14 days, the company ordered an additional 50TB of capacity to add to its Pillar Axiom. The ease of capacity scaling was a key benefit to SETAO.