Testing an SRM Information Provider
Contents
Introduction
To test the information published from an SRM we need to provide a Grid Information System which includes the test SRM. Unfortunately this requires at least three computers to properly test.
- A generic Grid UI node.
- A top level Grid Information Service (GIIS).
- The SRM to test.
The LHC Grid uses the LDAP protocol as the main information system for the Grid. This system is based upon a hierarchy of LDAP proxies that publish query LDAP servers and the republish this information. These services are often called BDII, which uses DBM as a persistence store for LDAP objects. BDII was developed in Perl from an original implementation in sh.
Configuring your test setup
Configuring the SRM
This should be done just fine by Yaim, but since I have to make YAIM do this correctly I am documenting this process here as well.
Configuring the GIIS
The Generic Information Provider (GIS) is a simple application designed to make configuring information systems relatively straight forward.
| Application | Path | RPM | Example |
|---|---|---|---|
| bdii | /opt/bdii/sbin/bdii | bdii-3.5.4-1_sl3 | /etc/init.d/bdii restart |
It is configured with two files.
/opt/bdii/etc/bdii.conf /opt/bdii/etc/bdii-update.conf
My example values are included here for reference.
/opt/bdii/etc/bdii.conf
BDII_PORT_READ=2170 BDII_PORTS_WRITE="2171 2172" BDII_USER=bdiiuser BDII_BIND=mds-vo-name=local,o=grid BDII_PASSWD=secret BDII_SEARCH_FILTER='*' BDII_SEARCH_TIMEOUT=30 BDII_BREATHE_TIME=60 BDII_AUTO_UPDATE=no BDII_AUTO_MODIFY=no BDII_DIR=/opt/bdii BDII_UPDATE_URL=http:// BDII_UPDATE_LDIF=http:// SLAPD=/usr/sbin/slapd SLAPADD=/usr/sbin/slapadd
/opt/bdii/etc/bdii-update.conf
SiteA ldap://elder.esc.rl.ac.uk:2135/mds-vo-name=local,o=grid T1 ldap://pnfs.gridpp.rl.ac.uk:2135/mds-vo-name=local,o=grid #SiteB ldap://pnfs.gridpp.rl.ac.uk:2135/mds-vo-name=local,o=grid #SiteB ldap://site-bdii.gridpp.rl.ac.uk:2170/mds-vo-name=RAL-LCG2,o=grid CERN-CIC ldap://lxn1194.cern.ch:2170/mds-vo-name=CERN-CIC,o=grid CERN-PROD ldap://prod-bdii.cern.ch:2170/mds-vo-name=CERN-PROD,o=grid CERN-SC ldap://lxb2088.cern.ch:2170/mds-vo-name=CERN-SC,o=grid
Please note that the last three lines include resources such as the replica catalogue so that lcg-tr works just fine.
The following two commands can be used to start and stop BDII. This should be done every time you change the configuration files.
/etc/init.d/bdii start /etc/init.d/bdii stop
BDII is non standard in its logging behaviour and logs its errors to /opt/bdii/var/tmp/stderr.log which should not contain lots of skipping lines. If it does you may need to hack a script /opt/bdii/sbin/bdii-update. Here is Our hack which is a little evil.
@@ -275,8 +275,7 @@
chomp(my $bad = $attr);
$attr =~ s/,\s+/,/g;
- #if ($attr =~ m/[:\s,]$bdii_bind\s*$/i) {
- if (1 == 2) {
+ if ($attr =~ m/[:\s,]$bdii_bind\s*$/i) {
#
# looks like recursive inclusion --> skip and warn!
#
What this patch does is prevent BDII being over cautious and skipping all LDAP queries which may be recursive, this is generally a good idea but unfortunately requires a second BDII to make things work with out this patch.
Configuring the UI
The UI node should already have all the correct software installed. We will need the following applications.
| Application | Path | RPM | Example |
|---|---|---|---|
| lcg-cr | /opt/lcg/bin/lcg-cr | lcg_util-1.3.5-1_sl3 | lcg-cr -v --vo dteam file:/etc/group -d elder.esc.rl.ac.uk |
| lcg-infosites | /opt/lcg/bin/lcg-infosites | lcg-info-api-ldap-2.5-1 | lcg-infosites --vo dteam all |
The following recipe works for me, where beech.esc.rl.ac.uk is the information system
$export LCG_GFAL_INFOSYS=beech.esc.rl.ac.uk:2170
$lcg-infosites --vo dteam all
****************************************************************
These are the related data for dteam: (in terms of queues and CPUs)
****************************************************************
#CPU Free Total Jobs Running Waiting ComputingElement
----------------------------------------------------------
2 2 0 0 0 lxn1184.cern.ch:2119/jobmanager-lcgpbs-dteam
2766 217 0 0 0 ce101.cern.ch:2119/jobmanager-lcglsf-grid_dteam
2766 215 0 0 0 ce102.cern.ch:2119/jobmanager-lcglsf-grid_dteam
**************************************************************
These are the related data for dteam: (in terms of SE)
**************************************************************
Avail Space(Kb) Used Space(Kb) Type SEs
----------------------------------------------------------
1 1 n.a elder.esc.rl.ac.uk
-579780983 2679021943 n.a dcache.gridpp.rl.ac.uk
265394264 437792936 n.a lxn1183.cern.ch
1000000000000 500000000000 n.a castorgrid.cern.ch
1000000000000 500000000000 n.a castorsrm.cern.ch
1000000000000 500000000000 n.a castorgridsc.cern.ch
If the output is somewhat similar to this the client and the GIS and the SRM's listed are correctly configured. Since this is unlikely it is a good idea to run the following command to establish if anything is coming to the lcg-infosites application. I typically use the following test.
$echo $LCG_GFAL_INFOSYS beech.esc.rl.ac.uk:2170 $ldapsearch -x -H ldap://$LCG_GFAL_INFOSYS -b 'Mds-vo-name=local,o=Grid' version: 2 # # filter: (objectclass=*) # requesting: ALL # ********************* SNIPPED DUE TO LENGTH ********************* GlueSchemaVersionMajor: 1 GlueSchemaVersionMinor: 1 # search result search: 2 result: 0 Success # numResponses: 280 # numEntries: 279
This application returns reasonable errors, so makes a good basic test to see if the new top level GIS is working. If the GIS is properly configured you should now be able to execute lcg-cr.
$lcg-cr -v --vo dteam file:/etc/group -d elder.esc.rl.ac.uk
Using grid catalog type: lfc
Using grid catalog : lfc-dteam.cern.ch
Source URL: file:/etc/group
File size: 557
VO name: dteam
Destination specified: elder.esc.rl.ac.uk
Destination URL for copy: gsiftp://elder.esc.rl.ac.uk:2811//pnfs/esc.rl.ac.uk/data/dteam/generated/2006-03-06/fileb9ff1ddb-5e5b-4424-94e4-60a139261988
# streams: 1
# set timeout to 0 seconds
Alias registered in Catalog: lfn:/grid/dteam/generated/2006-03-06/file-b45d2f19-a32b-43bc-82eb-140883f3bb55
557 bytes 0.32 KB/sec avg 0.32 KB/sec inst
Transfer took 3030 ms
Destination URL registered in Catalog: srm://elder.esc.rl.ac.uk/pnfs/esc.rl.ac.uk/data/dteam/generated/2006-03-06/fileb9ff1ddb-5e5b-4424-94e4-60a139261988
guid:abf6146f-a7a5-4f33-ae27-620900ec291f
If this worked you have successfully configured your top level information system. If this did not work please check that the URL starting with "file:/" contains only a single "/".
When I tried this at DESY for the first time I got this error,
$lcg-cr -v --vo dteam file:/etc/group -d nixon.desy.de Using grid catalog type: lfc LFC endpoint not found Using grid catalog : (null) LFC endpoint not found lcg_cr: Invalid argument
To resolve this error you must set the LFC_HOST variable
export LFC_HOST=prod-lfc-shared-central.cern.ch
