Download - Globus activities within INFN
Globus activities within INFN WP “Installation and Evaluation of the Globus
Toolkit” of the INFN-GRID Project Goal: evaluate the Globus toolkit as a GRID
framework providing basic services Which services can be useful ? What is necessary to integrate/modify ? What is missing ?
Duration: 6 months Results of this first evaluation used to plan future
activities
Tasks Security Information Service Resource Manager Globus deployment Data Access and Migration Fault Monitoring Execution Environment Management
Globus installed on ~ 35 machines in 11 sites
TORINO
PADOVA
BARI
PALERMO
FIRENZE
PAVIA
MILANO
GENOVA
NAPOLI
CAGLIARI
TRIESTE
ROMA
PISA
L’AQUILA
CATANIA
BOLOGNA
UDINE
TRENTO
PERUGIA
LNF
LNGS
SASSARI
LECCE
LNS
LNL
SALERNO
COSENZA
S.Piero
FERRARA
PARMACNAF
Status
ROMA2
Security (GSI) Already done:
Evaluation of the Globus security architecture
We like the general architecture, but: Granting local "identities" based only on certificate
subjects allows the existence of multiple valid certificates for the same subject
Authentication library not in sync with OpenSSL development
Cryptic diagnostics (e.g. "certificate chain too long" when the CA policy check fails)
Globus certificates (for hosts and users) signed by INFN certification authority
Security (GSI) To do:
Definition and implementation of architecture of CAs
Up to task force of the DataGrid project Make certificate requests easier Periodic update of CRL “Management” of grid-mapfile updates
I.e.: a certain Globus resource must be available to all members of a specific physics group
Information Service (GIS) Already done:
INFN MDS server serving Globus 1.1.1 and 1.1.2 installations
Lot of problems using the “default” American MDS server
Definition and implementation of test architecture of GIS (for Globus 1.1.3)
Web interface for browsing
Dc=bo, Dc=infn,dc=it,o=grid
Bologna
GIIS
INFN ATLAS GIIS
GIIS
Dc=mi,Dc=infn,dc=it,o=grid
Exp=atlas, o=grid
Top Level INFN GIIS
Dc=infn,dc=it,o=grid
Milano
GIS Architecture (test phase)
GRIS
ImplementedImplemented using INFNGRID distribution
To be implemented
Information Service (GIS) To do:
Netscape LDAP server as Top level INFN GIIS Tests on performance and scalability
Results used to define and implement the GIS architecture
Review the information gathered from the various machines and published in the GIS
Other tools and interfaces for Grid users and administrators
Resource Management (GRAM) Already done:
Job submission tests using Globus tools (globusrun, globus-job-run, globus-job-submit)
GRAM as uniform interface to different underlying resource management systems (LSF, Condor, PBS)
Some bugs found and fixed Standard output and error for vanilla Condor jobs globus-job-status …
Some bugs can be solved without major re-design and/or re-implementation:
For LSF the RSL parameter (count=x) is translated into: bsub –n x … Should be: bsub … x times
… Two major problems:
Scalability Fault tolerance
Globus GRAM Architecture
Client
LSF/ Condor/ PBS/ …
Globus front-end machine
Jobmanager
Job
pc1% globusrun –b –r pc2.pd.infn.it/jobmanager-xyz \ –f file.rsl
file.rsl:&(executable=/diskCms/startcmsim.sh)(stdin=/diskCms/PythiaOut/filename(stdout=/diskCms/Cmsim/filename)(count=1)
pc1 pc2
Scalability One jobmanager for each globusrun If I want to submit 1000 jobs ???
1000 globusrun 1000 jobmanagers running in the front-end machine !!!
%globusrun –b –r pc2.infn.it/jobmanager-xyz –f file.rslfile.rsl:
&(executable=/diskCms/startcmsim.sh)(stdin=/diskCms/PythiaOut/filename)(stdout=/diskCms/CmsimOut/filename)(count=1000)
It is not possible to specify in the RSL file 1000 different input files and 1000 different output files …
$(Process) in Condor Problems with job monitoring (globus-job-status) Therefore (count=x) with x>1 not very useful !
Fault tolerance The jobmanager is not persistent If the jobmanager can’t be contacted,
Globus assumes that the job(s) has been completed
Example of problem Submission of n jobs on a cluster managed
by a local resource management systems Reboot of the front end machine The jobmanager(s) doesn’t restart
Orphan jobs Globus assumes that the jobs have been successfully completed
Resource Management (GRAM) Already done:
Submission of Condor jobs to Globus resources (Condor-G and GlideIn mechanisms)
Evaluation of RSL as uniform language to specify resources The RSL syntax model seems suitable to define even
complicated resource specification expressions The common set of RSL attributes is often not sufficient
The attributes not belonging to the common set are ignored More flexibility is required
Resource administrators should be allowed to define new attributes and users should be allowed to use them in resource specification expressions (Condor Class-Ads model)
Same language to describe the offered resources and the requested resources (Condor Class-Ads model) seems a better approach
Resource Management (GRAM) Already done:
“Cooperation” between GRAM and GIS The information on characteristics and status of
local resources and on jobs is not enough As local resources we must consider Farms and not
the single workstations Other information (i.e. total and available CPU
power) needed The default schema must be integrated with other
info provided by the underlying resource management systems or by specific agents
GRAM & Condor & GIS
GRAM & LSF & GIS
Must be fixed
Jobs & GIS Info on Globus jobs published in the GIS:
User Subject of certificate Local user name
RSL string Globus job id LSF/Condor/… job id Status: Run/Pending/…
Resource Management (GRAM) To do:
Tests with GRAM API Tests with real applications and real environments
(CMS fall production) Already started
Memory leak in the job manager ?!?!?!? Solve the problems Identity a set of useful attributes of a Condor pool,
LSF cluster, PBS cluster that should be reported to the GIS, and integrate the default schema
Let’s start with information provided by the underlying resource management system
Second step: specific agents
GRID Globus deployment
Tools to enable local administrators to deploy the GRID software (now Globus 1.1.3 and related packages: OpenLDAP, …) Reduce complexity and manpower necessary
for installation Decrease errors during installations Collect bug fixes Include INFN customizations
Certificates (for hosts and users) signed by INFN CA … but user certificates signed by Globus CA are
accepted as well Preliminary architecture for GIS
GRID First step (July 2000)
Software distribution available on AFS Fixes for bugs found during first Globus
evaluations included INFNGRID installation guide
Instructions for INFN customizations included
Scripts to make certain steps (i.e. post-install operations) automatic
GRID Second step (now)
Pre-compiled distribution (available now for Linux Red Hat 6.1): INFNGRID 1.1
Script for installation and deployment: infngrid-install
Users decide to use INFN customizations or “standard” setup
Would you like the INFN setup (Y/N) ?
(1) Copy INFNGRID tar files from /afs/infn.it/project/infngrid/1.1/Linux to download dir(2) Decompress and untar INFNGRID distribution files in install dir(3) Configure INFNGRID software(4) Globus Setup(5) Configure GRAM services(6) Globus local deploy(7) GIIS Configuration====================================================
Condor and LSF
GRID Second step
Script for post install operations: globus-root-setup
Installation instructions for special environments (configuration of client machines, shared install-directory) included
List of included bug fixes Status
Tests performed in different environments (INFN, CERN, FNAL) “Officially” released Available to DATAGRID partners
(1) Modify system files and reactive the inetd daemon(2) Change owner to root of certain files for tighter security (3) Modify system wide login files (4) Start/restart Globus now(5) Configure gsi-wuftpd and restart the inetd daemon
GRID Next steps
Configuration of PBS as local resource management systems: 1.2
Support for Solaris 2.6: 1.2 We don’t plan (at least now) to support other platforms
Improvement of current no-precompiled distribution Eventual use of infngrid-istall script for both pre-
compiled and non pre-compiled distribution “Unattended” installation Management of updates Inclusion of GDMP: 1.2 Inclusion of other GRID software packages ?? Other works will be “triggered” by local administrators
and users
Data Management Already done:
Preliminary tests with GASS and gsiftp To do:
Tests with GlobusFTP and Replica Catalog Software (Globus Data Grid Alpha Release 2)
GARA
Preliminary tests considering both network and CPU advance reservation
Client
GARA API
GARA Network
Resource Manager
sunlab3 sunlab2
CISCO 7500 CISCO 7200
VC 100 MbpsFE FE
Server
Other tasks Fault Monitoring (HBM)
Evaluation of HBM for fault detection (for “system” and “user” processes)
Data collectors (implementing automatic recovery mechanisms)
… but the HBM package is not seeing active development
Execution Environment Management (GEM) Evaluation of GEM as service for code migration … but the GEM service now provides only limited
capabilities (executable staging)
Other info http://www.pd.infn.it/~sgaravat/ INFN-GRID/Globus