Storage Upgrade

Upgrade

/twiki/bin/viewfile/Main/EntryDescriptionNo77?rev=1;filename=sprace_distribuicao.jpg

May be some general guidelines:

• First start configuring a chimera namespace in our old server, and see that it works- See that it is working fine.
• Migrate to chimera names space on PnfsManager , from the old server to the new one, copying each pnfsid - See that it is working fine.
• Attach a new pool setup and start migration from old to new pools.
• Detach our old pool setup.

Guidelines for PnfsManager

The idea is have a single host for all the various dCache postgresql databases (both namespace, srm, pin manager and space manager) and we will run the PnfsManager with the chimera backend on the same host.

It should be great have a second and identical host as a warm standby, using continuous WAL archiving of postgresql. That way you will have a copy of all databases which are almost up to date (within a few minutes).

Following FNAL dCache configuration, pay attention to make a independent disk layout for databases, logs and system. An extreme performance tip is set all login parameter in pnfsSetup to 0 (no logging) and redirect all log output from the PNFS server to /dev/null. Also raid10 on database disk should be good.

Postgres

cd /tmp/
wget http://yum.pgsqlrpms.org/reporpms/8.3/pgdg-centos-8.3-4.noarch.rpm
rpm -ivh pgdg-centos-8.3-4.noarch.rpm
yum install postgresql postgresql-server
chkconfig postgresql on
service postgresql initdb
chkconfig postgresql on && service postgresql start
/usr/bin/postgres --version

Java

chmod +x jdk-6u12-linux-x64-rpm.bin 
./jdk-6u12-linux-x64-rpm.bin 
java -version

PNFS (Not necessary after when you change to chimera in our new server)

It is know a dependency with an libpq.so.4 for 32 bits. The first workaround is include a symbolic link of /usr/lib64/libpq.so.4 to /usr/lib/libpq.so.4 . The second one is install pnfs-postgresql using rpm -ivh --nodeps . Following some advices from FNAL presentation it is a common mistake put gridftp doors on admin node.

Also, we are migrating our data, it means that we want to migrate our PNFs to a different host, so we need to export our database. In our SOURCE admin it is done like a

/opt/d-cache/bin/dcache stop
/opt/pnfs/bin/pnfs stop
pg_dump -U pnfsserver admin >admin.dump 
pg_dump -U pnfsserver data1 >data1.dump

now inject it in your fresh install DEST admin node (we assume that all configuration files are set up)

dropdb -U pnfsserver admin
dropdb -U pnfsserver data1
createdb -U postgres admin
createdb -U postgres data1
psql -U postgres admin
and restart our pnfs system.
 Chimera 
 

 Start to  change to a Chimera , following these instructions
 
/twiki/bin/view/Main/DCche%2bChimera

to configure a new chimera instance, after shutdown pnfs and dCache.

http://trac.dcache.org/projects/dcache/wiki/pnfsDump2MigratePnfs2Chimera

To migrate, look also the attached Hands-on notes.

 Guidelines for SRMManager 
 

 Install only Java in this node.
 
 This will need a host certificate installed (copied from osg-se.sprace.org.br ) at /etc/grid-security and also a symbolic link to certificates directory. Use automount to mount our computer element directories:
 
 echo "/osgce     /etc/auto.spraid01   --timeout=30" >> /etc/auto.master
 echo "OSG     -rw,soft,bg,rsize=8192,wsize=8192,tcp   192.168.1.150:/OSG" > /etc/auto.osgce
 chkconfig autofs on
 ln -s /osgce/OSG /OSG
 ln -s /osgce/globus/TRUSTED_CA  /etc/grid-security/certificates 
 

 Install PhEDEx
 
https://twiki.cern.ch/twiki/bin/view/CMS/PhedexAdminDocsInstallation

 Pools 

To change our pools to a 64 bits node host, important directories: 

 The pool directory. It contains a control/ directory with meta data, a data/ directory with data, a one or more setup files.
 
 An entry in the pool_name_n.poollist (i.e. spraid02.poollist) file stored in the config/ directory of the dCache installation. This entry contains the path to the pool directory, the name of the pool and various settings.
 

Summarizing, supposing that we have the old and new pools in our setup configured:
1)copy data to the new pools
cd /raid2
rsync --rsh=ssh -av control root@spraid02_new:/diska/migration/
rsync --rsh=ssh -av data root@spraid02_new:/diska/migration/

2)Stop the dcache in your new server and move your data
/opt/d-cache/bin/dcache stop
cd /diska/migration/data
for ff in `ls -U`; do mv --reply=no $ff ../../data/ ; done
cd /diska/migration/control
for ff in `ls -U`; do mv --reply=no $ff ../../control/ ; done
/opt/d-cache/bin/dcache start

3)use admin interface to register these new data
ssh -c blowfish -p 22223 -l admin localhost
cd spraid01_new_1
pnfs register
..
logoff

In the same interface you can unregister the old pool
ssh -c blowfish -p 22223 -l admin localhost
cd spraid01_1
pnfs unregister
..
logoff

and stop watching it with 
cd collector@httpdDomain
unwatch spraid01_1
..
logoff

and stop dCache in your old instance /opt/d-cache/bin/dcache stop

Important: Do not start the pool at SOURCE after you moved the files. 


Alternatively we can bring the two new pool nodes online in the existing setup with new pool names, and the use the migration command of the pools to move the data - this command is available from dCache 1.9.1 and onwards.
migration move new_pool1

should do the trick. Run "help migration copy" for a description of all 
the options.











 Updates 
 Fulano em dd/mm/aaaa 
Coloca o que fez.
 Ciclano em dd/mm/aaaa 
Mais comentarios





-- MarcoAndreFerreiraDias - 23 Feb 2009