ELFms Meeting, Tuesday, 26th July, 2005
Present: Hugo Cacote, German Cancio, Alasdair Earl, Jan van Eldik, Thorsten
Kleinwort (minutes), Veronique Lefebure,
Marco-Emilio Poleggi, Dinesh
Sarode, Matthias
Schroeder, Miroslav Siket, Karol Stanislawek, William Tomlin
Last Minutes:
To be finished by Vlado
New Lemon API (Karol):
Karol gave an overview about the work that will be done to implement a new
Lemon API.
- GCM: The new API has to be backward compatible for all the software that
is using the current API, including sensorAlarm, CMDaemon, lemon-cli.
- Possibly, the clients that use the old API could be modified to use the
new API as well.
- WT: 'Connection Pooling' should be a requirement for the new API.
- MS: A new requirements document will be circulated in a few weeks from
now.
ELFms TWIKI:
German presented the new TWIKI web pages for ELFms:
https://uimon.cern.ch/twiki/bin/view/ELFms/WebHome.
- The idea is to provide an area for collaboration information.
- It should not replace existing web or the need for man pages.
- For the moment, four categories are set up and are available from the
man page.
- Authentication is done with AFS username and password.
- Hugo will help with setting up access control to the different areas
- Miroslav asked to have moderators for the different parts, German agreed
to enter them
- Tim Bell is supposed to give a presentation on the FIO TWIKI pages and
its usage in the next FIO Group Meeting, which should help people to
understand the usage of TWIKI. (The next group meeting is forseen for
September)
- Veronique requested to put change notification into place. German will
investigate
- Veronique asked if this is also the place to put the SMOD information.
It was agreed that a SMOD TWIKI is possibly a good idea, and this will be
discussed in the next SMOD meeting.
New set_partitions()
German explained that we had two different set_partition() functions in CDB
before, which were merged due to Miroslavs work into one function. But this
function is still referring to a schema that was set up before the official
schema that is now in the Quattor release.
- GCM: If we change, we should change all clients and tables and use only
the released Quattor schema.
- The question is, can this schema be extended to cope with RAID or LVM?
The existing schema in Quattor does not handle these cases yet
- Requirements for LVM and RAID should be gathered from the Section that
request/use it: FIO/DS (Jan van Eldik) and ADC/LE (Jarek Polok)
- The task to rewrite set_partitions() was on Miroslav, which he has done,
but GCM and TK were objecting that it is automatically Miroslav who will
look into the extension as well: to be discussed offline
- The next steps should be to get the requirements and propose a schema.
AOB:
- What is the current status of the new alarm system/ SURE phase out
- GCM: Current status: Two Lemon/Laser meetings have been held. The current
situation is that Laser cannot handle the large scale we have in Lemon
(#metrics times #hosts). A possible solution is to group individual alarms
on clusters/subclusters instead of individual nodes, which could reduce
the amount of data by a factor of 10-20. Another difference between the
current Laser implementation and Lemon is the number of reconfiguration
per time. While the Laser configuration is almost static (changes on a
monthly basis), the Lemon configuration changes on a daily bases, and this
includes
adding new or modifying existing alarms.
- The next step will be to have a prototype of
the Lemon
Alarm Gateway (LAG) integrated with LASER and see if the problems can be solved.
- For any of the options it seems to be the case that the current SURE
system will not be phased out within the next six month.
- What is the status of the RDBMS investigations?
- WT: A requirements document is out since some time, but not much
feedback was received so far. Apparently everybody is very busy and is
considering this as a low level task.
- Bill is still working on a design document, and Veronique is working
on a prototype for handling the RPM configuration