Difference between revisions of "Globus Online testing"
Line 1: | Line 1: | ||
This page is a summary of the tests that Karolis Eigelis from the EGI User Community Support Team (UCST) carried out with the European Globus Online service during October-November 2012. | This page is a summary of the tests that Karolis Eigelis from the EGI User Community Support Team (UCST) carried out with the European Globus Online service during October-November 2012. | ||
The tests helped UCST establish a [[Globus_Online_cookbook_for_EGI_VOs Globus Online Cookbook for EGI Virtual Organisations]] document. Other conclusions of the tests are: | The tests helped UCST establish a [[Globus_Online_cookbook_for_EGI_VOs Globus|Online Cookbook for EGI Virtual Organisations]] document. Other conclusions of the tests are: | ||
* Because of lack of regulations, restrictive permissions are set on local storage elements for the main SRM directory. | * Because of lack of regulations, restrictive permissions are set on local storage elements for the main SRM directory. |
Revision as of 10:25, 6 November 2012
This page is a summary of the tests that Karolis Eigelis from the EGI User Community Support Team (UCST) carried out with the European Globus Online service during October-November 2012.
The tests helped UCST establish a Online Cookbook for EGI Virtual Organisations document. Other conclusions of the tests are:
- Because of lack of regulations, restrictive permissions are set on local storage elements for the main SRM directory.
- VOInfoPath in BDII may be the path only for SRM interface, but patgh for the related GridFTP server is different.
- The SRM hostname that is advertised in BDII is not always the same site where the GridFTP server operates (but usually is). Globus Online uses GridFTP endpoints.
-bash-3.2$ lcg-info --list-se --bdii ldap://bdii.grid.sara.nl:2170 --vo dteam --query 'SE=bostorm.bo.infn.it' --attrs 'VOInfoPath' - SE: bostorm.bo.infn.it - VOInfoPath /dteam
VOInfoPath is /dteam for SRM interface:
-bash-3.2$ srmls -2 srm://bostorm.bo.infn.it:8444/dteam Picked up _JAVA_OPTIONS: -Xmx512M 0 /dteam/ 599 /dteam/20080528-143657-3356-0.txt 755 /dteam/20081117-145803-15525-0.txt 599 /dteam/20080528-143353-2676-0.txt 0 /dteam/misurelli 0 /dteam/testas 0 /dteam/-acls 599 /dteam/20080528-145209-4387-0.txt 755 /dteam/20081117-145800-15187-0.txt 0 /dteam/karolis 599 /dteam/20080528-144905-4048-0.txt 755 /dteam/20081117-150109-16203-0.txt 0 /dteam/generated 0 /dteam/dbauer
but if we connnect using GridFTP, /dteam directory even does not exist:
-bash-3.2$ uberftp bostorm.bo.infn.it "ls /dteam" 220 bostorm.bo.infn.it GridFTP Server 6.10 (gcc64, 1334324800-83) [Globus Toolkit 5.2.0] ready. 230 User dteam028 logged in. Could not list /dteam: 500-Command failed. : System error in stat: No such file or directory 500-A system call failed: No such file or directory 500 End.
because , it is a "symlink" in SRM WS to real directory which is /storage/dteam, but only SRM interface may access it:
-bash-3.2$ uberftp bostorm.bo.infn.it "ls /storage" 220 bostorm.bo.infn.it GridFTP Server 6.10 (gcc64, 1334324800-83) [Globus Toolkit 5.2.0] ready. 230 User dteam028 logged in. drwxrwxr-x 21 storm storm 4096 Sep 26 12:45 alice drwxrwxr-x 4 storm storm 4096 Sep 26 12:45 atlas drwxrwxr-x 3 storm storm 4096 Sep 26 12:45 cms drwxr-x--- 2 storm storm 4096 Sep 26 12:45 comput-er.it drwxrwxr-x 2 storm storm 4096 Jun 29 09:34 computer drwxrwxr-x 6 storm storm 4096 Oct 29 15:01 dteam drwxr-x--- 2 storm storm 4096 Sep 26 12:45 infngrid drwx------ 2 root root 16384 Jun 7 18:43 lost+found drwxrwxr-x 4 storm storm 69632 Nov 2 09:58 ops -bash-3.2$ uberftp bostorm.bo.infn.it "ls /storage/dteam" 220 bostorm.bo.infn.it GridFTP Server 6.10 (gcc64, 1334324800-83) [Globus Toolkit 5.2.0] ready. 230 User dteam028 logged in. Could not list /storage/dteam: 500-Command failed. : System error in scandir: Permission denied 500-A system call failed: Permission denied 500 End.
Directory listing on GlobusOnline
On GO Web, if directory is empty, GO Web fails with error listing that directory, if at least on file exists it works fine.
e.g. if i try to enter the directory which is empty i get:
"Could not access /dpm/nikhef.nl/home/dteam/karolis. Please check the path and try again. Error listing directory '/dpm/nikhef.nl/home/dteam/karolis/' on endpoint 'ekarolis#nikhef_DPM': Error: Embedded '/' in '/dpm/nikhef.nl/home/dteam/karolis' dirlist: dirlist/recurse.cpp:206: void _process_direntry(MLSDEntry*, OpDirlist*): Assertion `0' failed. hide details"
It fails on Nikhef DPM node, but works fine on SARA dCache node.
SRM -> GO endpoint
Go endpoint can be created from all three types (DPM, dCache, STORM) of storage elements within EGI, but SEs have to meet some requirements:
Software versions:
- GridFTP compatible versions with GlobusOnline access methods within EGI
Gathering info within EGI:
Retrieve all SEs from BDII
-bash-3.2$ ldapsearch -LLL -h bdii.grid.sara.nl -p 2170 -x -b mds-vo-name=local,o=grid '(objectClass=GlueSE)' GlueSEUniqueID|grep GlueSEUniqueID:|awk '{print $2}' > EGI_ALL_SE_HOSTS
Count how much hosts we have
bash-3.2$ wc -l EGI_ALL_SE_HOSTS 489 EGI_ALL_SE_HOSTS
OR
-bash-3.2$ lcg-info --list-se --bdii ldap://bdii.grid.sara.nl:2170 --attrs 'SE'|tr -d '\-SE:[:blank:]'|tr -s '\r' '\n'|sort|uniq|wc -l 483
Probe for GridFTP 2811 port EGI_ALL_SE_HOSTS
-bash-3.2$ for i in `cat EGI_ALL_SE_HOSTS`;do echo ¨Connecting to $i `nc -w 2 $i 2811`¨;done|tee EGI_ALL_SE_GRIDFTP_VERSIONS
Count how many of them responded with GridFTP open
-bash-3.2$ grep FTP EGI_ALL_SE_GRIDFTP_VERSIONS|wc -l 299
What are the GridFTP Server versions
-bash-3.2$ awk '{print $8}' EGI_ALL_SE_GRIDFTP_VERSIONS|sort|uniq 2.3 2.7 2.8 3.23 3.28 3.33 6.10 6.14 6.5
Lets see how many types of storage exists in EGI
-bash-3.2$ ldapsearch -LLL -h bdii.grid.sara.nl -p 2170 -x -b mds-vo-name=local,o=grid '(&(objectClass=GlueSE))' GlueSEImplementationName|grep ^GlueSE|sort|uniq GlueSEImplementationName: ARC GlueSEImplementationName: bestman GlueSEImplementationName: BeStMan GlueSEImplementationName: CASTOR GlueSEImplementationName: classicSE GlueSEImplementationName: dcache GlueSEImplementationName: dCache GlueSEImplementationName: DPM GlueSEImplementationName: HDFS GlueSEImplementationName: StoRM GlueSEImplementationName: xrootd
How many of DPM, dCache, Storm ?
-bash-3.2$ ldapsearch -LLL -h bdii.grid.sara.nl -p 2170 -x -b mds-vo-name=local,o=grid '(&(objectClass=GlueSE)(GlueSEImplementationName=dcache))' GlueSEUniqueID|grep ^GlueSE|wc -l 79 -bash-3.2$ ldapsearch -LLL -h bdii.grid.sara.nl -p 2170 -x -b mds-vo-name=local,o=grid '(&(objectClass=GlueSE)(GlueSEImplementationName=dpm))' GlueSEUniqueID|grep ^GlueSE|wc -l 217 -bash-3.2$ ldapsearch -LLL -h bdii.grid.sara.nl -p 2170 -x -b mds-vo-name=local,o=grid '(&(objectClass=GlueSE)(GlueSEImplementationName=storm))' GlueSEUniqueID|grep ^GlueSE|wc -l 54
Retrieve only SRM hosts from BDII
bash-3.2$ ldapsearch -LLL -h bdii.grid.sara.nl -p 2170 -x -b mds-vo-name=local,o=grid '(&(objectclass=GlueService)(GlueServiceType=SRM))' GlueServiceEndpoint |grep GlueServiceEndpoint| awk '{print $2}'|sort|uniq|sed -e '/managerv1/d'|sed -e 's/httpg:\/\///g'|sed -e 's/:.*//' > EGI_ALL_SRM_ENDPOINTS
Count how much SRM endpoints we have
-bash-3.2$ wc -l EGI_ALL_SRM_ENDPOINTS 423 EGI_ALL_SRM_ENDPOINTS
OR
-bash-3.2$ lcg-info --list-service --bdii ldap://bdii.grid.sara.nl:2170 --query 'ServiceType=SRM'|awk {'print $3'}|sort|uniq|sed -e '/managerv1/d'|sed -e 's/httpg:\/\///g'|sed -e 's/:.*//'|wc -l 420
Configuration issues:
- GridFTP must accept VOMS proxy e.g. user is mapped to /home/dteam009 at storage element
- Main storage directory used by SRM must allow VO member using his VOMS proxy to access it e.g. /storage/dteam, and not only his home directory which is e.g. /home/dteam009
- BDII or SRM WS should provide the real path to the main storage directory used on storage element e.g. VOInfoPath from BDII MAY contain the path for SRM ONLY and SRM does the mapping to the real path, but GridFTP does not know how to do that mapping.
Note: Configuration issues may be different, depending on how EGI decides to configure SRM/GridFTP/SEs in general and how GlobusOnline finds important information within EGI):
File operations
All operations below were performed succesufully:
a) copy file using SRM and listing file using GO Web
b) copy file using GO Web and listing file using SRM
c) checking if correct Disk Used Sizes are reported by BDII (transfering at least 1GB)
d) deleting file using SRM (file was put by using GO Web)
e) deleting file using GO Web (file was put by using SRM)
f) Transfer file from NIKHEF DPM to SARA dCache.
1) Picked up few SEs:
DPM 1.8.2 - tbn18.nikhef.nl
dCache 2.2.4 - srm.grid.sara.nl
STORM 1.8.2 - bostorm.bo.infn.it
2) Finding SRM endpoints for these SEs:
-bash-3.2$ lcg-info --list-service --bdii ldap://bdii.grid.sara.nl:2170 --vo dteam --query 'ServiceType=SRM' --attrs 'ServiceEndpoint'| grep tbn18.nikhef.nl - Service: httpg://tbn18.nikhef.nl:8446/srm/managerv2 - ServiceEndpoint httpg://tbn18.nikhef.nl:8446/srm/managerv2 -bash-3.2$ lcg-info --list-service --bdii ldap://bdii.grid.sara.nl:2170 --vo dteam --query 'ServiceType=SRM' --attrs 'ServiceEndpoint'| grep srm.grid.sara.nl - Service: httpg://srm.grid.sara.nl:8443/srm/managerv2 - ServiceEndpoint httpg://srm.grid.sara.nl:8443/srm/managerv2 -bash-3.2$ lcg-info --list-service --bdii ldap://bdii.grid.sara.nl:2170 --vo dteam --query 'ServiceType=SRM' --attrs 'ServiceEndpoint'| grep bostorm.bo.infn.it - Service: httpg://bostorm.bo.infn.it:8444/srm/managerv2 - ServiceEndpoint httpg://bostorm.bo.infn.it:8444/srm/managerv2
3) Finding SRM VO home path for these SRM endpoints:
-bash-3.2$ lcg-info --list-se --bdii ldap://bdii.grid.sara.nl:2170 --vo dteam --query 'SE=tbn18.nikhef.nl' --attrs 'VOInfoPath' - SE: tbn18.nikhef.nl - VOInfoPath /dpm/nikhef.nl/home/dteam -bash-3.2$ lcg-info --list-se --bdii ldap://bdii.grid.sara.nl:2170 --vo dteam --query 'SE=srm.grid.sara.nl' --attrs 'VOInfoPath' - SE: srm.grid.sara.nl - VOInfoPath /pnfs/grid.sara.nl/data/dteam -bash-3.2$ lcg-info --list-se --bdii ldap://bdii.grid.sara.nl:2170 --vo dteam --query 'SE=bostorm.bo.infn.it' --attrs 'VOInfoPath' - SE: bostorm.bo.infn.it - VOInfoPath /dteam
4) Listing files in SRM endpoints:
-bash-3.2$ srmls -2 srm://tbn18.nikhef.nl:8446/dpm/nikhef.nl/home/dteam Picked up _JAVA_OPTIONS: -Xmx512M 0 /dpm/nikhef.nl/home/dteam/ < output truncated > -bash-3.2$ srmls -2 srm://srm.grid.sara.nl:8443/pnfs/grid.sara.nl/data/dteam Picked up _JAVA_OPTIONS: -Xmx512M 512 /pnfs/grid.sara.nl/data/dteam/ < output truncated > -bash-3.2$ srmls -2 srm://bostorm.bo.infn.it:8444/dteam Picked up _JAVA_OPTIONS: -Xmx512M 0 /dteam/ < output truncated >
5) Listing used and total space on SEs (in GB units):
-bash-3.2$ ldapsearch -LLL -h tbn18.nikhef.nl -p 2170 -x -b mds-vo-name=resource,o=grid '(&(objectClass=GlueSA)(GlueSAAccessControlBaseRule=VO:dteam))' GlueSATotalOnlineSize GlueSAUsedOnlineSize dn: GlueSALocalID=test:replica:online,GlueSEUniqueID=tbn18.nikhef.nl,Mds-Vo-na me=resource,o=grid GlueSATotalOnlineSize: 32 GlueSAUsedOnlineSize: 4 -bash-3.2$ lcg-info --list-se --bdii ldap://bdii.grid.sara.nl:2170 --vo dteam --query 'SAVOs=VO:dteam' --attrs 'UsedOnline,TotalOnline'|grep srm.grid.sara.nl -A 7 - SE: srm.grid.sara.nl - UsedOnline 0 0 0 - TotalOnline 0 15000 58681 Note: three values because it is dCache. -bash-3.2$ lcg-info --list-se --bdii ldap://bdii.grid.sara.nl:2170 --vo dteam --query 'SAVOs=VO:dteam' --attrs 'UsedOnline,TotalOnline'|grep bostorm.bo.infn.it -A 3 - SE: bostorm.bo.infn.it - UsedOnline 0 - TotalOnline 100
6) Uploading file from UI via SRM client to SE endpoint:
-bash-3.2$ srmcp -2 file:///100M.rand srm://tbn18.nikhef.nl:8446/dpm/nikhef.nl/home/dteam/100M.rand -debug Picked up _JAVA_OPTIONS: -Xmx512M Storage Resource Manager (SRM) Client version 2.1.5-16 Tue Oct 30 09:24:36 CET 2012: In SRMClient ExpectedName: host Tue Oct 30 09:24:36 CET 2012: SRMClient(https,srm/managerv2,true) SRMClieldapsearch -LLL -h tbn18.nikhef.nl -p 2170 -x -b mds-vo-name=resource,o=grid '(&(objectClass=GlueSA)(GlueSAAccessControlBaseRule=VO:dteam))' GlueSATotalOnlineSize GlueSAUsedOnlineSizentV2 : user credentials are: /O=dutchgrid/O=users/O=egi/CN=Karolis Eigelis SRMClientV2 : connecting to srm at httpg://tbn18.nikhef.nl:8446/srm/managerv2 SRMClientV2 : srmPrepareToPut , contacting service httpg://tbn18.nikhef.nl:8446/srm/managerv2 copy_jobs is empty Tue Oct 30 09:24:39 CET 2012: srm returned requestToken = bd7521b6-65ab-4c21-a301-9c484ef81098 Tue Oct 30 09:24:39 CET 2012: sleeping 5 seconds ... SRMClientV2 : srmStatusOfPutRequest , contacting service httpg://tbn18.nikhef.nl:8446/srm/managerv2 Tue Oct 30 09:24:44 CET 2012: no more pending transfers, breaking the loop copy_jobs is not empty copying CopyJob, source = file:///100M.rand destination = gsiftp://hooiwagen.nikhef.nl/hooiwagen.nikhef.nl:/export/data/dteam/dteam/2012-10-30/100M.rand.160938141.0 GridftpClient: memory buffer size is set to 131072 GridftpClient: connecting to hooiwagen.nikhef.nl on port 2811 GridftpClient: gridFTPClient tcp buffer size is set to 1048576 GridftpClient: gridFTPWrite started, source file is java.io.RandomAccessFile@74bfed5a destination path is hooiwagen.nikhef.nl:/export/data/dteam/dteam/2012-10-30/100M.rand.160938141.0 GridftpClient: gridFTPWrite started, destination path is hooiwagen.nikhef.nl:/export/data/dteam/dteam/2012-10-30/100M.rand.160938141.0 GridftpClient: set local data channel authentication mode to None GridftpClient: parallelism: 10 GridftpClient: waiting for completion of transfer GridftpClient: starting a transfer to hooiwagen.nikhef.nl:/export/data/dteam/dteam/2012-10-30/100M.rand.160938141.0 GridftpClient: DiskDataSink.close() called GridftpClient: gridFTPWrite() wrote 104857600bytes GridftpClient: closing client : org.globus.ftp.GridFTPClient@3b5b123b GridftpClient: closed client execution of CopyJob, source = file:///100M.rand destination = gsiftp://hooiwagen.nikhef.nl/hooiwagen.nikhef.nl:/export/data/dteam/dteam/2012-10-30/100M.rand.160938141.0 completed SRMClientV2 : srmPutDone , contacting service httpg://tbn18.nikhef.nl:8446/srm/managerv2 srmPutDone status code=SRM_SUCCESS copy_jobs is empty stopping copier
Note: In the log above we see that in fact the real SE endpoint is "hooiwagen.nikhef.nl", which is not advertized in BDII, but pre-configured inside SRM WS.
7) Listing file via GO Web and deleting the file via GO Web.
-bash-3.2$ srmls -2 srm://tbn18.nikhef.nl:8446/dpm/nikhef.nl/home/dteam/100M.rand Picked up _JAVA_OPTIONS: -Xmx512M 104857600 /dpm/nikhef.nl/home/dteam/100M.rand -bash-3.2$ srmls -2 srm://tbn18.nikhef.nl:8446/dpm/nikhef.nl/home/dteam/100M.rand Picked up _JAVA_OPTIONS: -Xmx512M Tue Oct 30 09:32:12 CET 2012: Return status: - Status code: SRM_FAILURE - Explanation: null SRM_INVALID_PATH File/directory 0 /dpm/nikhef.nl/home/dteam/100M.rand does not exist.
8) Uploading file from GO Web to SE endpoint, listing with SRM client and deleting with SRM:
-bash-3.2$ srmls -2 srm://tbn18.nikhef.nl:8446/dpm/nikhef.nl/home/dteam/1G.rand Picked up _JAVA_OPTIONS: -Xmx512M 1048576000 /dpm/nikhef.nl/home/dteam/1G.rand -bash-3.2$ srmrm -2 srm://tbn18.nikhef.nl:8446/dpm/nikhef.nl/home/dteam/1G.rand Picked up _JAVA_OPTIONS: -Xmx512M -bash-3.2$
9) Uploading 5GB size file from laptop to SE using GO Web and checking if BDII is reporting space correctly.
Before upload:
-bash-3.2$ ldapsearch -LLL -h tbn18.nikhef.nl -p 2170 -x -b mds-vo-name=resource,o=grid '(&(objectClass=GlueSA)(GlueSAAccessControlBaseRule=VO:dteam))' GlueSATotalOnlineSize GlueSAUsedOnlineSize dn: GlueSALocalID=test:replica:online,GlueSEUniqueID=tbn18.nikhef.nl,Mds-Vo-na me=resource,o=grid GlueSATotalOnlineSize: 32 GlueSAUsedOnlineSize: 4
Listing file with SRM client:
-bash-3.2$ srmls -2 srm://tbn18.nikhef.nl:8446/dpm/nikhef.nl/home/dteam/karolisPicked up _JAVA_OPTIONS: -Xmx512M 0 /dpm/nikhef.nl/home/dteam/karolis/ 5242880000 /dpm/nikhef.nl/home/dteam/karolis/5G.rand
After upload:
-bash-3.2$ ldapsearch -LLL -h tbn18.nikhef.nl -p 2170 -x -b mds-vo-name=resource,o=grid '(&(objectClass=GlueSA)(GlueSAAccessControlBaseRule=VO:dteam))' GlueSATotalOnlineSize GlueSAUsedOnlineSize dn: GlueSALocalID=test:replica:online,GlueSEUniqueID=tbn18.nikhef.nl,Mds-Vo-na me=resource,o=grid GlueSATotalOnlineSize: 32 GlueSAUsedOnlineSize: 9
10) Making transfer from Nikhef DPM to SARA dCache using GO Web and listing file using SRM on dCache:
-bash-3.2$ srmls -2 srm://srm.grid.sara.nl:8443/pnfs/grid.sara.nl/data/dteam/karolis Picked up _JAVA_OPTIONS: -Xmx512M 512 /pnfs/grid.sara.nl/data/dteam/karolis/ 5242880000 /pnfs/grid.sara.nl/data/dteam/karolis/5G.rand
And notification by email was received:
=== Task Deta-bash-3.2$ srmls -2 srm://srm.grid.sara.nl:8443/pnfs/grid.sara.nl/data/dteam/karolis Picked up _JAVA_OPTIONS: -Xmx512M 512 /pnfs/grid.sara.nl/data/dteam/karolis/ 5242880000 /pnfs/grid.sara.nl/data/dteam/karolis/5G.rand ils === Task ID : 429d16a4-2278-11e2-bf56-1231380b8963 Task Type : TRANSFER Parent Task ID : n/a Status : SUCCEEDED Request Time : 2012-10-30 09:57:48Z Deadline : 2012-10-31 09:57:48Z Completion Time : 2012-10-30 09:58:56Z Total Tasks : 1 Tasks Successful : 1 Tasks Expired : 0 Tasks Canceled : 0 Tasks Failed : 0 Tasks Pending : 0 Tasks Retrying : 0 Command : API 0.10 eu Label : from NIKHEF DPM to SARA dCache Sync Level : n/a Data Encryption : No Checksum Verification: No Delete : No Files : 1 Files Skipped : 0 Directories : 0 Bytes Transferred : 5242880000 Bytes Checksummed : 0 MBits/sec : 626.016 Faults : 0