Availability computations with ACE

Europe/Amsterdam
Universe

Universe

Description
Participants: D. Collados, T. Ferrari, E. Imamagic, D. Zilaskos CREAM CE in availability statistics =================================== During the ongoing transition from LCG-CE to CREAM, it is important to remember that at the moment CREAM Nagios probe results are NOT included in availability computations. This has several implications. For example, if a site has a working CREAM instance and the LCG-CE instance is down being under migration to CREAM, but it is still marked production certified, then the availability statistics will be affected (as the working CREAM instance is not counted and the lcg-ce node will fail tests). So it is important that the status of lcg-ce node is changed in GOCDB before maintenance starts. IMPACT of CREAM CE tests on AVAILABILITY STATISTICS ================================================== CERN is carrying out a comparison of availability statistics between CREAM CE/ARC vs LCG-CE/ARC (comparisons are based on the OPS VO). The purpose of this is to verify if the conditions apply to decommission LCG-CE and replace it with CREAM. At the March GDB the results of the comparison will be reported (http://indico.cern.ch/conferenceDisplay.py?confId=126130). The comparison is based on ACE, which is capable of handling different profiles. ACE will be used for this comparison even if not officially in production (it is planned to be released during March 2011 if no software deficiencies are found in the meantime). EGI is interested in keeping LCG-CE in the OPS availability computations until it will be officially supported (i.e. until end of security support). Will this require a split of the OPS WLCG profile and of the OPS EGI profile? When WLCG will declare LCG-CE ready for replacement with CREAM, and if the comparison of availability will go well, then WLCG may consider the possibility to remove LCG-CE completely from the availability computations. However, before doing so, ACE needs to go to production, this will take a couple of months after release. In June 2011 conditions will be checked to decide if a split of the OPS profile is needed. MANAGEMENT OF PROFILES IN ACE ============================= This is done centrally at the moment. In case of intervention needed, a ticket can be opened to the SAM Nagios Support Unit (the 3rd level support one that will be soon implemented). In Summer 2011 a GUI will be available for direct editing of an own profile. The AUTHZ framework for this has still to be defined. MYEGI ===== Is there a possibility to extract availability statistics in a format that can be processed by Nagios probes? MyEGI exports - through its programmatic interface (web Services) - availability DB figures in json and xml format. A query can be built using this PI interface. Documentation is available. REPORT GENERATION OF MONTHLY AVAILABILITY STATISTICS ===================================================== A full document in confluence exists that describes how monthly availability league tables can be published (it is linked from the EGI wiki tools section). openreport generates the pdf file. openreport is used to 1. upload xml templates, 2. perform data connections 3. query the database 4. structure the pdf document The existing xml templates for WLCG could be easily customized for EGI, if this requires just light changes such as management of a new profile, and of new service types. This level of change could be managed by BARC for EGI on best effort. If a totally new type of record were needed, this would require a new xml template. Several approaches are possible 1) deploy an own DB and consume information through AMQ 2) use the existing DB A new DB can be setup for EGI if needed, this is simple especially if historical information does not need to be migrated. Today for ACE this database is only ORACLE. There are plans to release an ACE component for MySQL in a long term if manpower permits. Central databases at CERN will remain in Oracle. DECISION: for the time being the same reporting system will be used (WLCG and EGI). The possibility to have an own reporting system will be considered if conditions for this change arise.
The agenda of this meeting is empty