View
213
Download
0
Category
Preview:
Citation preview
Enabling Grids for E-sciencE
SA1
EGEE-II INFSO-RI-031688
The Pre-Production Service in WLCG/EGEE
A. Retico, N. Thackray CERN – Geneva, Switzerland
PPS contributes to the quality of gLite middleware:
- Software and release documentation
- e-science applications from VOs benefit of a
- Feedback is given for early bug fix to gLite before releasing into production
are validated through the operation in a real grid in real conditions
scaled environment for validation and debugging before they are moved into production
The PPS grid counts about 30 sites providing
resources and manpower
Pre-Production Service cooperates with the OSG Integration Testbed (ITB) to verify EGEE/OSG inter-operability across gLite releases
The EGEE
http://www.cern.ch/pps
The EGEE Pre-Production Service (PPS) is a distributed service whose goal is to give early access to new services to WLCG/EGEE users in order to evaluate new features and changes in the release
Enabling Grids for E-sciencE
SA1
EGEE-II INFSO-RI-031688
Pre-Production Service Infrastructure
A. Retico, N. Thackray CERN – Geneva, Switzerland
~30 sites~30 sites
~50 CEs~50 CEs
live on PPS BDII
~60 SEs~60 SEs
4 FTSs4 FTSs
16 countries16 countries
SLA for PPS sitesAll PPS sites must comply with all of the following terms, unless there is a specific agreement with the PPS management to the contrary.
1) The running of a PPS site will probably not take 1 FTE. However, during the time that a new release is passing through the pre-production phase, the site administrator(s) must make their PPS site their absolute top priority. Outside of this pre-production phase, the pre-production site can take a lower priority (but not 0).2) Must be at least 0.4 FTE (equivalent to 1 person, 2 full days per week).3) The site must send representation to all PPS operations meetings.4) The site must supply either ≥ 1 core service or CE + WNs + SRM storage + MonBox or all of these.
DILIGENT (http://www.diligentproject.org) pre-production infrastructure is composed by 6 sites: • CNR: Pisa, Italy • ENG: Rome, Italy • ESA: Rome, Italy • SNS: Pisa, Italy • UNIBAS: Basel, Suisse • UoA: Athens, Greece
The "diligent" VO is also supported by 10 other EGEE PPS sites. This infrastructure is mainly used to: • store of the DILIGENT user communities data • store the DILIGENT services archives used in the deployment of the DILIGENT service • execute the DILIGENT watermarking application • execute feature extraction applications • execute application defined as part of a compound service • manage the DILIGENT users and their groups/roles
DILIGENT (http://www.diligentproject.org) pre-production infrastructure is composed by 6 sites: • CNR: Pisa, Italy • ENG: Rome, Italy • ESA: Rome, Italy • SNS: Pisa, Italy • UNIBAS: Basel, Suisse • UoA: Athens, Greece
The "diligent" VO is also supported by 10 other EGEE PPS sites. This infrastructure is mainly used to: • store of the DILIGENT user communities data • store the DILIGENT services archives used in the deployment of the DILIGENT service • execute the DILIGENT watermarking application • execute feature extraction applications • execute application defined as part of a compound service • manage the DILIGENT users and their groups/roles
What about the HEP VOs?As a subjective and qualitative indicator of the effort spent on different activities done in PPS, the plot below shows the number of e-mail exchanged on dedicated mailing lists. • Standard Usage (5%): VOs use SW regularly released • Special Activities (37%): VOs test non-certified SW (e.g. SRMv2)• Release Testing (15%): few selected sites do pre-deployment testing• Operations (43%): ~ 30 sites maintain a service runningHigh operation costs compared to poor (standard) usage by VOs => Revision of the mandate in study
What about the HEP VOs?As a subjective and qualitative indicator of the effort spent on different activities done in PPS, the plot below shows the number of e-mail exchanged on dedicated mailing lists. • Standard Usage (5%): VOs use SW regularly released • Special Activities (37%): VOs test non-certified SW (e.g. SRMv2)• Release Testing (15%): few selected sites do pre-deployment testing• Operations (43%): ~ 30 sites maintain a service runningHigh operation costs compared to poor (standard) usage by VOs => Revision of the mandate in study
test report
Certification > PPS
PPS Coord pre-testing Team repository Mgr PPS SitesSA3
Kick-off
- Write Release Notes - Update repo@CERN
Consistency check of repo@CERN
pre-deployment testing
(installation, configuration.
SAM)
Prepare release bulletin
release notes
apt status report
test report
Deploy?
- Publish Release Bulletin
Synchronize repo@CNAF
Y
Handle roll-back
N
Upgrade
Broadcast
Open Bugs and update test reports
RELEASE to PROD
list of issues
to DEV
ELOPER
S
PPS > Production
PROD SitesPPS CoordSA3
tentative list
Create candidate list of patches
EMT meeting: finalize list of patches
meeting: Define list of “known issues”
final list
FEEDBAC
K from PPS
- Write Release Notes - Update Repository
final list
Verify status of associated bugs
final list
Prepare Broadcast
release notes
Upgrade
BROADCAST
Fri
Mon
Tue
Thu
Mon
Tue
… 2
weeks
later...
Wed
list of issues
Release Process
PPS sites contributing to release operations:
• PPS-CNAF: manages PPS repository
• PPS-LIP, CERN_PPS, KIAM-PPS, UKI-SOUTHGRID-BHAM-PPS, PPS-IFIC, CESGA-PPS, prague_cesnet_pps: pre-deployment testing
PPS sites contributing to release operations:
• PPS-CNAF: manages PPS repository
• PPS-LIP, CERN_PPS, KIAM-PPS, UKI-SOUTHGRID-BHAM-PPS, PPS-IFIC, CESGA-PPS, prague_cesnet_pps: pre-deployment testing
Tools for monitoring, operations and policy enforcement are in place in PPS in the same fashion as in production. SAM, gstat, GridView, FCR
CIC Portal is also extended to support pre-production operations
General operations supported by the EGEE ROCs
SAM Tests are run independently in PPS by PPS-RAL and PPS-CYFRONET (UPATRAS did it in the past)
All content available at
http://www.cern.ch/pps
Recommended