These are the installation procedures for a complete installation site LCG independent of any external resources (Deliverable A3.3 - 2005Q4).
Thanks to Sabah Salin and Alessandra Forti for support and discussions.
Hardware/Software architecture.
Savannah portal: http://savannah.cern.ch/ http://www.gridpp.ac.uk/deployment/links.html LCG rollout archives: http://www.listserv.rl.ac.uk/archives/lcg-rollout.html Download releases: http://lcg.web.cern.ch/LCG/Sites/releases.html Wiki: http://goc.grid.sinica.edu.tw/gocwiki/FrontPage LCG software is a set of specific functional computers running modules that interact with each other during the job submission, running and deliver processes. The processes are:CE and RB never could be installed in the same computer. There are installations where RB and BDII are together. These are the conputers with standard Scientific Linux 3.0.4 installation from www.scientificlinux.org. The log is available here.
- User interface (UI): It is the module that allows users interact with grid. It contains programs to submit jobs and commands to the grid, check job status, recover outputs, etc.
- Resource broker (RB): It is responsable to match the jobs with resources, implement policies, and send jobs for processing in a remote resource (Computer element).The definition of resources are written in the JDF file using ClassAds (see condor manual and LCG user manual for more details).
- Computer element (CE): it is the batch manager (PBS) that allocates resources in the farm to run users' software. There are interactive jobs, parallel jobs, and simple jobs.
- Worker Nodes (WN): it is the computer that will run your software. It contains the packages installed, libraries, and necessary local data for processing.
- Storage Element (SE): it is the mass storage element in grid.
- Berkeley Database (BDII): it is the catalog of resources published in the grid. It is distributed in two levels. The first catalog all CE BDII entries. The second, in the CE, contains the information about the farm managed by the CE.
- Proxy manager (PX): stores proxies for long time jobs in the grid.
IP Hostname nick Function Cert IS_ip IS_host IS_nick BDII PX MON * RB_ip RB_host RB_nick RB * CE_ip CE_host CE_nick CE * WN_ip WN_host WN_nick WN ... UI_ip UI_host UI_nick UI SE_IP SE_host SE_nick SE * * e-Science certificate are required in these machines The data server using NFS is cap.hep.man.ac.uk. Most of the current servers get delivered with two ethernet: eth0 should be connected to the public network (194.36.XX.XX) and eth1 to the private network (192.0.168.XX). The advantage is for example, when accessing data server (cap) will not share network connection with AFS. Another important point is: when writing data from your code, do not write records remotely (AFS, NFS, etc). Write outputs in a local disk, and at the end of your job copy it to a remote resource.
Verify if your certificate is installed:
At desktop pc73.hep.man.ac.uk, user jamwer: ls .globus usercert.pem userkey.pem If you do not have certificate, you have to obtain one before request hosts certificates.Run firefox browser:
firefox Site: https:/ca.grid-support.ac.uk accept this certificate permanently (or you will not be able to recover the host certificate) Request a certificate Server certificate DSN is the server hostname type host These are the outputs, to confirm the information: E-Mail jamwer2000@hotmail.com DNS Name host/IS_host L HEP OU Manchester Fully qualified domain name IS_host User Data Role User Registration Authority Manchester HEP NOTBEFORE Mon Nov 7 09:02:15 2005 UTC PIN 553848908d51c23446b713a058c2f91ed656927e PROFILE UKHOST RA Manchester HEP ROLE User SERIAL 362016 SUBJECT_ALT_NAME DNS: IS_host TYPE SPKACLater, you will receive an email such as:
From : ca@ca.grid-support.ac.uk Sent : 07 November 2005 16:23:11 To : jamwer2000@hotmail.com Subject : UK eScience CA - New Issued Certificate (Serial: 5141) Dear Customer, Your certificate with the serial number 5141 and the DN: emailAddress=jamwer2000@hotmail.com,CN=host/IS_host,L=HEP,OU=Manchester,O=eScience,C=UK has just been generated by the UK e-Science CA. You can import it directly into the same browser you requested the certificate from by following this link: https://ca.grid-support.ac.uk/cgi-bin/pub/pki?cmd=getcert&key=5141&type=CERTIFICATE You can also do this by going to https://ca.grid-support.ac.uk/pub and clicking on "Import Certificate into Browser" and entering the serial number: 5141Recovering the certificates:
Go in the same computer, account, and browser (firefox) https://ca.grid-support.ac.uk Import certificate into browser Serial number: (see number in the subject from ca email, in my case 5141) At firefox: Edit/Preferences/Advanced/Certificates/manage certificates Mark one certificate Backup type pem phrase WARNING: in the "Certificate backup password" and "...(again)" type any character and backspace (it will be empty!) Save as computer_name.p12Converting p12 to pem format
See http://www.hep.man.ac.uk/local/grid/grid-cert-FAQ.html openssl pkcs12 -in name.p12 -clcerts -nokeys -out namecert.pem Enter Import Password: Just type enter! MAC verified OK openssl pkcs12 -in name.p12 -nodes -nocerts -out namekey.pem Enter Import Password: Just type enter! MAC verified OKAt each host computer:
cd /etc/grid-security/ cp /nfs/work/users/jamwer/GRID/namecert.pem hostcert.pem cp /nfs/work/users/jamwer/GRID/namekey.pem hostkey.pem chmod 400 hostkey.pem chmod 444 hostcert.pem where XX is 30, 31, 32, and 40.The firewall issue.
Using grid from your laptop/desktop requires some ports are open in the firewall. The description about it are available at: LCG application, middleware & security - last 2 pages are important, and not able to be printed! Globus requirements Security and Firewall Settings Reporting on experiences operating Globus through firewall.
Top
|
|
|
Feedback to: jamwer@hep.man.ac.uk |