This wiki is obsolete, see the NorduGrid web pages for up to date information.
ARC Client Tools Transition
ARC Client Tools Transition from ng* to arc*
The purpose of this page is to document all differences between the ng* and arc* client tools.
Upgrade procedure
If you installed ARC 0.8.x client tools using yum or apt, it is advised to remove old installations to avoid possible confusion.
New ARC client tools are called arcsub, arcstat etc, and are almost identical in purpose and functionality to those called ngsub, ngstat etc., however, command line options have changed, and several new features are now available. The following documents describe installation and usage details:
Error messages in new client tools are still at times unpolished; please report all confusing messages to User Interface Bugzilla section.
Differences between ng* and arc*
| Function/option | ng* output | arc* output | Comment |
|---|---|---|---|
| Compute clients | |||
| -x and -X flags in the ng-client | -x has different meaning, anonymous queries are not used, -X (gsi-gssapi binds) not available | ||
| sync | ngsync always truncates the .ngjobs file, then fill it up with the jobs it found in the specified clusters | arcsync by default merges the existing jobs with the found ones, and only truncates, if -T is given | |
| specifying target clusters | ngresub -k specifies the target clusters | arcresub -k is for keep, and -q is for specifying the target clusters | |
| renew proxy | ngrenew refuses to renew the proxy if it's not older then the local proxy | arcrenew always renews | |
| cat -f | ngcat -f can follow the file (like tail -f) | no -f option in arccat | |
| Timeout (-t) | Timeout only for MDS query (should be double-checked) | Timeout is applied in each step (connect, close, send info) | Should be documented |
| Data clients | |||
| ls file:///tmp/file.txt | /tmp/file.txt | file.txt | ngls on gsiftp and http prints only filename, difference only for local files |
| ls -l file:///tmp/file.txt | /tmp/file.txt file 0 "Wed Feb 16 14:24:16 2011" * * | <Name> <Type> <Size> <Creation> <Validity> <CheckSum> <Latency>
file file 0 2011-02-16 14:24:16 (n/a) (n/a) (n/a) |
|
| ls -m /tmp/file.txt | accessperm:rw-------
atime:Wed Feb 16 14:24:16 2011 ctime:Wed Feb 16 14:24:16 2011 group:1073 mtime:Wed Feb 16 14:24:16 2011 owner:122831 path:/tmp/file.txt size:0 type:dir |
file
accessperm:rw------- atime:2011-02-16 14:24:16 ctime:2011-02-16 14:24:16 group:1073 mtime:2011-02-16 14:24:16 owner:122831 path:/tmp/file.txt size:0 type:file |
arcls prints file before metadata, ngls prints wrong type |
| cp -i | ngcp -i prints new line every second | arccp -i prints progress as status bar with no new-line when source file size is known (i.e., source file is a local file), otherwise progress is shown in numbers | |
| Stage files from tape to disk | ngstage | No equivalent arc* command | |
| 3rd party data transfer | ngtransfer | No equivalent arc* command | |
| Set Grid ACL for files | ngacl | No equivalent arc* command | |
| Test suite | |||
| List available resources | ngtest -R | arcinfo | arcinfo is a new tool, patially taking over ngtest functionalities |
| All clients | |||
| long options | prefixed with single dash | prefixed with double dash | |
Other
- list clusters from file (-C)
- arc* cannot get the list of clusters from a file
- list of GIISes from a file (-G)
- arc* cannot get the list of GIISes from a file
- Caching (cp -y)
- ngcp and arccp use different time formats, so cache from ngcp is incompatible with cache from arccp
Output file nested URL semantics
Previously an output file like lfc://srm://srm.org/data/test@lfc.org/grid/file1 would be treated as follows:
The file is uploaded to
srm://srm.org/data/test/file1
and this replica is registered under
lfc://lfc.org/grid/file1
In other words the last part of the logical name is appened to the physical name. In ARC 11.05 and later, the semantics were changed to allow a different logical and physical name, and so the whole physical filename must be specified in the inner part of the URL. If the previous example was used an 11.05 cluster would upload the file to
srm://srm.org/data/test
Note that in 11.05 and later the job description syntax allows an cleaner way to specify output files (and URL options), for example in xRSL:
("outputfiles" = ("myfile" "lfc://lfc.org/grid/file1" "location=srm://srm.org/data/test/file1" "spacetoken=MYTOKEN"))
