A summary of RMLL 2010

This year back to Bordeaux. I wasn’t there since 2001 IIRC. And my first presentation was only in 2006 in Nancy. I arrived on Tuesday, made all the registration process, and was able to attend the first sessions of the afternoon. Here are my notes and comments around the event. Most presentations are available from the link to the RMLL Web site.

I attended mainly the system administration track this year. What was interesting is the large interest of multiple projects to interface to the new FusionInventory Agent in order to benefit from iys ubiquity. This project looks very promising and is lead by clever and hard working leaders. GLPI, GOsa, Mondo Rescue, maybe LIuxCOE are good candidate for integration. We had long discussions between the various people involved in these projects, and the face to face interactions allowed us to make progresses that would not have been possible remotely. I really look forward working with them.

Some projects are also interested by the project-builder.org approach (which wasn’t presented in itself). Again GOsa and FusionInventory will have a closer look at it.

Jean-Pierre Laisné on 2020 FLOSS roadmap

Presentation is State of the Art for 2010. 2010-10-01 at OWF 3rd edition of the roadmap
Started in 2008. 40 contributors. Replacement of Gartner/IDC/… by the FLOSS Community itself.
Impact on R&D – Industry – Society
Martin Michlmayr is HP’s representative.

Reminder on 2009 (Sun/Oracle, Open Data, Open HW – Apple, …)
Comparison of FLOSS and Forests (bio-diversity)
Primary forests (gcc, linux, perl), cultivated forests (JBoss, MySQL, …), IKEAs (Red Hat, Canonical,…) National parcs ? Forest management ?
Recommendation on contribution and recognition as a public knowledge asset (common goods ?)
FLOSS as common (David Bollier) – BRIC Countries (Michael Tiemann) – Freedom in the Cloud (Philippe Aigrain) – Diaspora constitution.
Use of Co-ment to give feedbacks to the current proposals http://2020flossroadmap.org

Very interesting initiative, presented very clearly by Jean-Pierre. Needs contribution.

Boothes visit
Meeting with Mandriva contributors and Mandriva users association, FDN

With GOsa, GLPI and Linagora contributors. (I talked mainly with Benoit Mortier and David Durieux)

Sylvain Beucler on Savannah hosting platform

Lots of info on licensing and legal aspects (IP geolocalisation issues with US, DMCA, AGPLv3, …)
Machine under Coreboot + Xen. (1 server 3000 projects) – Another french server for Gna! (1300) projects. Use of NSS for users management.
Mirmon tool for mirror monitoring + Apache2::Geo:Mirror
Munin for monitoring – + AWstat and Webalizer (custom apache log to get real donwload size not size of file)

Olivier Thauvin on distrib-coffee mirror management

Interesting presentation of problems and issues related to managing a public ftp/http server.
What to mirror, where to mirror, to whom you provide access.
Second Mandriva Mirror WW. (1,2 TB for all Mandriva – 110 GB for one version vs 90 GB for fedora 12)
Problem of time and costs (Download 20 Mb/s mini – 100 Mb/s – Upload 100 Mb/s mini – 1 Gb/s)
High # of concurrent rather small read accesses – Only 80% of raw storage size at the end.
HW or SW Raid is mandatory. SATA is best compromise price/perf
distrib-coffee = 20 TB (13 kEUR) + server (1.5 kEUR)
Separated /var at least (I’d add /var/log as a separate FS)
A single /var/ftp FS (easy with LVM + for resize) with high number of inodes
ext3/4, xfs (btrfs not tried)
http: pb for # of connections + timeouts long => Adapt MaxClients + SymlinksIfOwnerMatch
pb of massdownloader => mod-limitipconn very efficient
ftp: good for it but disable recursive ls at / and allow passive ftp.
rsync: lots of advantages but heavy in I/O + mem. Disable compression (useless most of time) and checksums
bittorrent/smb (ro)/nfs (ro) these 2 internally
QoS: tc (very hard)/iptables. thb (bandwidth control) + pfifo and cfq
Mirrors: Use mirror brain to help mirror admins. Use rsync (-a -H)
Use script (lock + cron + timeout) => mmm (perl)

Good presenttion on a real case example of mirror management, that every sysadmin would find useful.

Bruno Cornec on MondoRescue

30+ people. Including grub2 developer, GLPI developer.
Hint to look at grub-mkrescue.

Jesse Vincent on RT

10 people in Best Practical Solutions, his firm.
RT is 14 years old !!
Keeps track of what is going on + metadata + business logic
Everything has: ID – timestamp – history should not be changed
Mindset: No task too small or too big – Update often – bad ticket better than no ticket. track discussions
Explained the configuration flexibility of the tool (roles, tasks nature, ACLs, customs fields, custom scripts, strong mail interaction, graphs, reports+cron …).
Pure perl dep 5.8.D3 only + API REST GET/POST
Overlays for code, webui, translation, …
up to 70000 tickets/day
Tons of plugins
Developed RTFM and RTIC

Had not enough time to cover all the material. Obviously an expert to who you could give hours of talk 😉

Nicolas Charles on CFengine 3

Normation society
Importance to manage configurations with tools (combination of SCM + distribution management)
Tool written in C by Mark Burgess
Automation and normalization. Convergence to SLA. Based on Promise theory + language associated.
Usage as standalone or as policy server.
Local agent responsible of local activities (scalability + network resiliency)
LDAP only on commercial version 😦
Windows native/FreeBSD package version on commercial version 😦
Some strange choices between the FLOSS/Commercial version.
Install under /var/cfengine !! Warning: Executes as root.
Explanation of the promise syntax
Competitors: puppet or bcfg2

Still not sure of Cfengine myself. Find it very complex to deal with.

With OpenSuSE and Mandriva contributors. (I talked mainly with Vincent Untz and Olivier Thauvin)

Walid Nouh on GLPI

Gestion Libre de Parc Informatique: FLOSS Management of large set of systems
Has become a large project (100kLOC+, 25 languages, an association behind Indepnet.net, large community 1M+ machines managed referenced)
Easy Deployment (Apache / PHP / MySQL deps + packages)- Web based for usage – plugin system (60 available) – Custom UI however fields are mandatory. Multi-entity management (0.70+).
Configuration is the most interesting (LDAP/AD/CAS integration)
OCS often coupled with GLPI to manage inventory (now evolving to FusionInventory) However DB is separate as GLPI processes the data and clean them.
Management of contracts (alerts on garanty limits/date limits in general, leasing/rental)
Helpdesk feature (used by the HPISC BTW). There are better tracking tools. The advantage here is the link with the suppliers/inventory/contract/…
Support ticket creation from mails/We interface. Rules engine now available by entities.
Multiple levels of interface by role/competencies. User (simplified interface)/Support Engineer (std interface). UI for reports and stats.
ITIL support added with migration process from the existing versions, and choice to use it or not by the users (so more parameters with good defaul
ts in the same version) Brought in 0.78 (not planed originally – RC2 2010-07-06 !! – asked by French CNAM)
New features in 0.78:
– automatic actions (cron like)
– real notification system (now by entity + mail templates + easy support for i18n)
– LDAP improvements (dynamic import of users)
– New management of dropdowns (titles)
– Helpdesk (ticket lifecycle, urgency and impact level – matrix to compute priority – priority for major incidents) – mail routing engine (SPOC)
Needs feedbacks on RC2 !
Plugins 0.72 not compatible with 0.78 (and no doc yet !) but lots of improvements
Some already ported on Wiki (reports, PDF, webservices)
GLPI 0.80 (soon after 0.78):
– ITIL v1 (SLA – User satisfaction)
– Links between tickets and with KB
– Multiple affectations of tickets
GLPI 0.81:
– Change management for ITIL

HackWeek in June 2010. Documentation is lagging. Even in french.
First Glossary definition. Then french doc (target 0.78). English doc (target 0.80)
GLPI / RT ? Separation of Helpdesk/Machines Management ?

Our HP/Intel Solution Center uses the tool already, but these new features are very promising, and the development around the tool active.

David Durieux & Gonéry le Boulder on FusionInventory

David: Siprossii – Gonéry: Atos
Single Agent multi-OS (Win/Mac/Unix/Linux/BSD) offering modularity (agent and server)
Union of GLPI tracker and SNMP query + OCS Unix Inventory Agent in perl
Features: Asset inventory local (HW, SW, VMs, …) and remote SNMP (Switches, printers, …)
Network discovery (nmap, netbios, SNMP)
Remote wake up of the agent now available. (HTTP server available in the agent). Push & Pull.
SSL support and p2p for sw deployment.
FusionInventory deployment module in plan (2010).
Agent Server communication compatible with OCS. Server Plugin for GLPI + PHP libfusion.
CPAN distribution.
Work also on passive network discovery, Backup solution, OPSI agent, monitoring.

Why an OCS fork ?
– No open decision process
– Very long standing bugs (64 bits support, UTF-8, no major change since 2006)
– Change request from dev rejected
– Server has serious drawkbacks (mod_perl, security, DB schema, link with GLPI, SOAP !!)

One of the most interesting session of the week. Definitely a project to follow. Lok forward the possibility to work with them.

Benoit Mortier on GOsa²

LDAP based administration platform for users, groups & systems
Linux & Windows
Homogeneous and nice Web interface.
Everything is stored in the LDAP server (including deployment info)
Deployment with FAI (Debian mainly) and OPSI (Windows) with gosa-si
Linkage with GLPI in study ?

GOsa is always a platform that would merit more noise around it as its LDAP centric view is definitely a very good approach. And Benoit is an excellent speaker.

Bruno Cornec on LinuxCOE

~30 people attended. Including FusionInventory leaders.

Sébastien Delcroix on WinPackage

Developed for CITIC74 (IT center for public sector + all schools of the department)
Development of their own deployment solution to take in account the large matrix of deployment (found no tool to do it)
Need of meta-packages => apt-get for Windows.
Managing conflicts, deps, mandatory interactivity.
Used python (py2exe to make executables and pywin32) + XML.
Called Winpackage.
Future: package signature and integrity
Need also a client / server tool for management and reporting (WWMSD).
Same technical env.
To be done: reporting, SSL support + certificates.
Not published yet, waiting for security to be finished.
Demo of installation of thunderbird with:
winpackage install -v thunderbird
winpackage upgrade (updates to 3.0.4)
400 pkgs existing with 3600 machines

Need to wiat till publication to see really what it brngs. Also in competition with the following.

Fabrice Flore-Thebault on Wpkg

Work based on comparison between desired state and current state + report.
Simple project but may not be suited for large companies.
File repo = CIFS – Javascript + XML
dependency management through profiles.
there is a fusion inventory patch (ask Gonéry 😉
Feedback on customer cases: 100/200 machines – ~20 pkgs
Installed in 1 day – project mode approach – simple to manage, easy to transfer to helpdesk.
Demo of package installation – more windows look and feel.

Seems that the presentation underestimated the usage of wpkg, and its ability to fit most needs. A candidate for project-builder.org then.

Benoit Mortier on OPSI

Manage windows desktop and servers – Python + java – GPL – 12 years old
Installation from scratch – A la kickstart
HW inventory, SW Inventory – SW deployment – Multi repository – SW license management – JSON interface – CLI interface including under Windows – Java graphical interface
Scripting language to build deployment packages – cpio format
Actions on group of machines dynamically generated (dep on previous inventory)
Creates a pcpatch user (with appropriate rights) managing the installations – uses a CIFS share.
Updates done either by blocking user login or after user connection (+timeout)
DOesn’t modify Windows for install. Just integrates and automates.
Interfaces with driverpack for HW support at install time.
Rewrite back-end to manage more than 1500 machines (RDBM schema optimisation)
New groups nearer from profile and policies
Slow links support (delta transfer)
FusionInventory and GLPI integration.

Is OPSI the most best kept secret of the Munich migration program ?

Libre Dinner
With Benoit Mortier, Philippe Capdepuy and Daniel Leprince. Not as well organized as previous years, but always a “must be” place to meet with 200+ FLOSS people at once !


Jean-Philippe Faye – Nelly Mahu-Lasson – Walid Nouh
Use of a similar tool for all 150 org in CNAMTS (from 100 to 1000 people) using multiple entry channels (phone, mail, …)
Automation of escalation and report – Single DB of knowledge
Use ITIL – But delegated administration => Better costs management, better governance
Currently there is duplication of incidents report in different DB
Next step is a have a single portal (internal dev) coupled to GLPI with a single DB between orgs at level 1 (long term at level 2 as well). Goal to
manage SLA
Also need for precise inventory (170000+ systems).
OTRS (more ITIL, german) vs GLPI (more present in the french community, roadmap interesting)
10 orgs representative define the specs, qualify the product and ensure that changes occurs.
Another team is responsible of the product evolution and coordination (Budget, resources and risks).
They have L3 capabilities + commit rights.
In CNAMTS context the main issue was that GLPI was single entity. Request to the community leaders to make adaptations, and commitment to reverse the modifications in the project.
Features were discussed, agreed upon between customer and project.
Project was sceptic as many promises were made, without having seen achievement.
Walid Nouh was one external consultant used to made GLPI adaptations, but was external to the project. Slowly he became GLPI commiter, patch after patch
Features developed (multi entities, delegated admin, recursivity, dictionaries, LDAP connectors, improved rules engine, …)
Plugins developed to manage transtition from previous inventories (reversed), Pulse2 comm, SOAP/XML-RPC, OCS sync, reports, uninstall, …
Reims primary center (4 people originaly, now 2 + Walid all members of the community on 6 total) – Referenced users

Rule = no fork so need to align with project roadmaps (e.g. version 0.78 unplanned was added to support the customer needs). Step 1 operational (20
10-06-15 – 2010-07-15) genreralisation Oct/Nov 2010.
Next step already planned (lifecycle, SLA, stats, …)

During round table:
What about another big customer ?
Play on paramters, community remains in charge.

Importance of support contract.
COTS vs custom development. FLOSS is not only specific dev. It’s also COTS.
This system has not yet evolved to create followers.

Ademe has developed tooling for that.
DIfficult to get financial feedback due to lack of historical data.

An excellent presentation (also longer) on a real testimony of collaboration between a large pulic sector customer and a FLOS project. Proof of success is the generalization put in place today.

Patrice Bertrand on FLOSS strategy for enterprises

CEO of SMILE – Representative of CNLL
Presentation of a simple Governance approach for enterprises.
Mentioned letter from the DoD CIO explaining a FLOSS orientation.
Short due to time constraints (had slides on licenses, policies, …)

Philippe Montarges on Open Innovation

Alter Way – OWF
Definition of Open Innovation: Opened R&D to the outside (partners, external collaborators, …) but result can be fully proprietary ! This open is
NOT the open of the Open Source !
People think that FLOSS similar to generic medecine creation. Which is reducing its value.
4 approaches:
– Shared R&D (LF, Mozilla)
– Spinout (Eclipse, Jikes)
– Selling Complements (dual licensing, Websphere, Ingres, Aladdin)
– Donated components (Symbian)

=> aproach FLOSS/OI, FLOSS/notOI, notFLOSS/OI, notFLOSS/notOI
(Oracle was missing to the picture ! HP as well ! Red Hat as well !)
No mention of cloud either 😦
Goal was to justify that FLOSS is the main laboratory of OI, and is an economy in itself.
Short presentation of OWF.

Not a very interesting presentation IMHO, as it didn’t entered in concrete details at all. Maybe more for CxO type of attendance.

François Letellier on FLOSS is central to ICT Innovation

87% of top500 projects on SF.net are not innovative.
But same % in proprietary world !!
E.g. France on patents and publications is at or below the average.
Doesn’t count FLOSS e.g. Measure of innovation on what you can count (researchers, start-ups, patents, publications, …)
But that doesn’t work for FLOSS. Counting contributors is hard, innovation is collective and incremental (lots of steps).
“The finger pointing to the moon is not the moon”
Simulation show that extreme protection are counter-productive.
FLOSS allows 36% of R&D savings at EU level. (FLOSS would cost 12 EUR B to redevelop)
Ex: Ingres (BSD) passed by Relational Technologies/ASK/CA is now again FLOSS.
(Reference given to the book of Henry Chesbrough on Open Innovation)
FLOSS is central to ICT innovation, is OI in the SW area, provides multiple business models.

François is a lively speaker having an excellent view on his domain.

Round table:
Bruno asked about Open Financement for OI ?
Branding is important for investors.

Tags: , , , ,

2 Responses to “A summary of RMLL 2010”

  1. elvir Says:

    Thank you Bruno

  2. Back to RMLL 5 years after ! | Bruno Cornec's Blog Says:

    […] So this year I’ll give a status of where we are with the MondoRescue project for my tenth anniversary of project leadership and after the last presentation I did 5 years ago. […]

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: