ELFms/LEMON meeting, 22/01/2004

Present: Helge Meinhard, Piotr Kolet, Marek Misiowiec, Dennis Waldron, Miroslaw Siket, Manuel Guijarro, Harry Renshall, German Cancio (minutes), Fabio Trevisani, Hugo Cacote, Jan van Eldik

PSF - Python Scripting Framework

• Marek’s slides can be found here.
• PSF consists of two parts:
• a framework for easing writing MSA sensors, shielding the MSA/sensor protocol details. Implementation is done using Python
• Linode: a collection of C libraries for monitoring the Linux installation servers at CERN
• Marek has choosen Python as implementation language due to the large number of libraries available, and its (true) object orientation. The framework is compact (some 1K LOC). How much experience does FIO have with Python?
• With PSF, we have three MSA sensor frameworks: the other ones are written in C++ (Sylvain) and Perl (JanvE).
• Linode is a use-case targeted application, but some of the code could be reused for other purposes. It does remote monitoring of the installation servers but not provide any generic remote monitoring functionality. We could reuse some of the ideas/concepts for remote service monitoring (eg. are all neccessary servers reachable from a given rack’s network segment)
• The framework has not yet been released to other users. Marek will leave CERN in 10 weeks, what is the support level by ADC? (German to check)
• Subject to the above - should we integrate the Linode metrics into the FIO repository?

CCS prototype 2

• Fabio: the kernel on the PVSS servers was upgraded and the machines rebooted.
• HMS integration (automated adding/removing of machines into the PVSS configuration): no progress.

Oracle MR

• David’s report can be found here.
• The current procedure for restarting OraMon on the production server is considered insufficient (periodically checking if OraMon still works..). It is decided that a ‘daemon dead’ alarm will be sent to the operators, which will inform us via the mailing list (Jan van Eldik to follow up).

MSA / sensors

• Jan: Repackaging of MSA and sensors under progress – the MSA RPM should come from the EDG autobuild, an additional RPM provides CERN-specific sensors and configurations. Jan will deploy the MR API RPM onto LXPLUS.
• Hugo: how to store switch sensor measurements? Helge: in I/O per time unit, like KB/s.
• Currently, our software is spread over three CVS servers: PVSS repository (on Helges server), EDG repository (in Lyon), and FIO repository (on the CERN central CVS servers). As the EDG repository will be frozen, it will be neccessary to move the repository module containing the Lemon sources to CERN. At the same time, the monolithic source structure and build makefiles should be revisited. Bill has proposed some FIO-wide guidelines for structuring CVS repositories. German suggests using the quattor build tools for package generation. Manuel raises the point that PS should have write access to this new repository. German to check and propose a new CVS layout.

Derived metrics

• Dennis: working on adding wall clock time restrictions to Cmsensor. The XML parsing has been improved (now using Apache Xerces), as well as the MSA parameter handling and debugging instrumentation.

AOB

• It is decided to track actions via an action list (see below).
• For the next meeting, German asks for a ‘to do’ list from all project participants, listing the items which to their understanding need to be addressed in 2004. The collected items would be used as input for a detailed planning in 2004.
• Examples:
• CDB integration of MSA, OraMon, AlarmDisplay configurations
• Integrating the new MR server (S_Server) into OraMon
• Software re-packaging, in particular sensors
• OraMon stability, including fail-over / redundancy techniques
• AlarmDisplay GUI using WebStart
• MR_ API improvements (eg. convert from SOAP lists to arrays)
• Providing MR access to experiment users
• Deploying Cmsensor locally and globally
• Local Recovery actuator framework
• ...

Action Items:

New actions:

· [A1] 22/1/04 All: Provide list of 'TO DO' items to German for 2004 planning

· [A2] 22/1/04 German: Check with ADC support level for PSF framework

· [A3] 22/1/04 JanvE: DAEMON_DEAD alarm when production OraMon daemon dies.

· [A4] 22/1/04 JanvE: Deploy the MR API RPM onto LXPLUS

· [A5] 22/1/04 German: Propose a new CVS layout (check with Bill Tomlin)