Symbolic links in iRODS

320 views
Skip to first unread message

eric liao

unread,
Nov 26, 2009, 12:41:55 PM11/26/09
to irod-chat
Hi,

I was wondering if it is possible to make symbolic links in iRODS? Could I register data that already exist in iRODS similar to how 'ireg' works for local files?
This would be very useful when I want to share access to data in a private collection without needing to provide access to that collection or needing to copy the data to a 'public' space.
I guess this is conceptually similar to a 'ticket-based' access system but with no time out for the ticket.

Cheers,
Eric

mw...@diceresearch.org

unread,
Dec 1, 2009, 1:04:56 PM12/1/09
to irod...@googlegroups.com
Hello Eric,

We are thinking about providing symbolic link for collections. Would that be sufficient ?

Currently, you can sort of do it using the imcoll command, but it is kind of dangerous
and not recommended. With the current iRods, suppose I have a collection
/tempZone/home/mwan/testdir. ils -L of this collection shows:
 ils -L testdir
/tempZone/home/mwan/testdir:   
  mwan              2 demoResc                     4698 2009-04-22.16:53 & foo3
        /data/mwan/rods/RODS/Vault/home/mwan/testdir/foo3   

The file foo3 is stored as a UNIX file
 /data/mwan/rods/RODS/Vault/home/mwan/testdir/foo3

I can the do the following:
imkdir /tempZone/home/mwan/testdir1
imcoll -m filesystem /data/mwan/rods/RODS/Vault/home/mwan/testdir /tempZone/home/mwan/testdir1

So I have just linked testdir1 to testdir

 ils -l testdir1
/tempZone/home/mwan/testdir1:
  mwan         mountP demoResc                     4698 2009-04-22.16:13 & foo3


So, now foo3 is visible under testdir1.

The danger is that if I modify foo3 through testdir1, the new size and checksum values will
not be reflected in the metadata creating inconsistency.


Mike






--
"iRODS community is contagious - don't you know enthusiasm is contagious".
 
iROD-Chat web page: http://groups.google.com/group/iROD-Chat

Terrell Russell

unread,
Dec 1, 2009, 1:55:57 PM12/1/09
to irod...@googlegroups.com
Mike, this is interesting.

I have one question about the data displayed - foo3 seems to have two
different timestamps in your example, with the 'second' file having an
earlier timestamp. Can you explain why they are different?


Thanks,

Terrell

mw...@diceresearch.org

unread,
Dec 1, 2009, 5:27:31 PM12/1/09
to irod...@googlegroups.com
Terrell,

This icat does not keep metadata of files in a mounted collection. So the timestamp of foo3
via ils testdir1 (a mounted collection) is based on the timestamp of the UNIX file.
When ils is through testdir (a mormal irods collection), the timestamp came from the icat.
foo3 actually has multiple replica and I left that out to avoid confusion. So the timestamp
is based on one of these copies.
Reply all
Reply to author
Forward
0 new messages