On Mar 14, 2014, at 7:54, LEIBOVICI Thomas <[email protected]> wrote:
> OK, I got it.
>
> Actually, robinhood considers any file modification or change as an access,
> assuming that if a user changes file data or metadata recently, he still
> cares about the file.
> Thus, rbh actually takes access time = MAX(atime, mtime, ctime),
> so it is perfectly expected that
> rbh's
> last access >= atime.
>
Does that mean the policy I'm using is unnecessarily redundant?
Policy default
{
condition { last_access > 15d and last_mod > 15d }
}
I set it that way on the off chance that a user deliberately modified a file's
atime to be older than the mtime.
thanks,
-Peter
-----
Peter Doherty |
Research Systems Administrator | Harvard Medical School
HMS Research Computing | https://rc.hms.harvard.edu/
> I first understood in your initial mail that rbh's last access was outdated,
> i.e. < actual atime.
> This is not expected, this is why I cared. But it finally doesn't seem to be
> the case.
>
> Regards,
> Thomas
>
> On 03/14/14 09:30, [email protected] wrote:
>> Hi Thomas,
>>
>> This is my policy:
>> Policy default
>> { condition { last_access > 3d } }
>>
>> I ran robinhood with the dry-run option for my tests, this is a line from
>> the log:
>>
>>
>>
>>
>> 2014/03/13 11:00:12 robinhood@xxx[6649/6]: Purged
>> '/scratch/xxx/xxxx/fichmeg/petit/m1.bin' using policy
>> 'default', last access
>> 20.8d ago | size=1048576, last_access=1392909857
>> ,
>> last_mod=1392312225, osts=ost#6: 39, ost#8: 39, ost#1: 80,
>> ost#3: 48, ost#7: 39, ost#9: 39, ost#0: 80, ost#2: 48
>>
>>
>>
>>
>>
>> This last access is corresponding to
>> 20/2/2014 à 16:24:17 but you can see
>> with stat command that the real last access was on 13/02/2014
>> and the change time was 20/02/2014:
>>
>>
>> # stat /scratch/xxx/xxxx/fichmeg/petit/m1.bin
>>
>>
>> File: « /scratch/xxxx/xxxx/fichmeg/petit/m1.bin »
>>
>>
>> Size: 1048576 Blocks: 2048 IO Block: 4194304 fichier
>>
>>
>> Device: 2c54f966h/743766374d Inode: 144115205272502310 Links:
>> 1
>>
>>
>> Access: (0644/-rw-r--r--) Uid: ( 211/ xxxx) Gid: ( xxx/ xxxxx)
>>
>>
>>
>> Access: 2014-02-13 18:23:45.000000000 +0100
>>
>> Modify: 2014-02-13 18:23:45.000000000 +0100
>>
>>
>>
>> Change: 2014-02-20 16:24:17.000000000 +0100
>>
>>
>>
>>
>> But, despite this, it's working. I change the acess time on
>> one file and run an another test with robinhood:
>>
>>
>>
>>
>>
>> # stat /scratch/xxxx/xxxx/gig2.bin
>>
>>
>> File: « /scratch/xxxxxxx/gig2.bin »
>>
>>
>> Size: 4294967296 Blocks: 8388640 IO Block: 4194304 fichier
>>
>>
>> Device: 2c54f966h/743766374d Inode: 144115205255725241 Links:
>> 1
>>
>>
>> Access: (0644/-rw-r--r--) Uid: ( 211/ xxx) Gid: ( xxx/ xxx)
>>
>>
>>
>> Access: 2014-03-13 10:59:22.000000000 +0100
>>
>> Modify: 2014-02-11 13:42:44.000000000 +0100
>>
>>
>>
>> Change: 2014-02-20 16:24:17.000000000 +0100
>>
>>
>>
>> Then I run "robinhood -S -O --dry-run" and "robinhood -O --dry-run" and when
>> i check the log, this file was not purged. So it's ok for me. I was just
>> surprised by the date on the log.
>>
>> About GPFS, we have two different environments, one with Lustre Filesystem
>> and an another with GPFS Filesystem. As we want to use robinhood to manage
>> file on Lustre, we'd like to manage GPFS files also with robinhood.
>> I'll give you some feedback if i can implement this.
>>
>> Have a good day and thank you,
>>
>> Regards,
>>
>> Emilie
>>
>> De: "LEIBOVICI Thomas" <[email protected]>
>> À: [email protected]
>> Cc: [email protected]
>> Envoyé: Jeudi 13 Mars 2014 14:42:26
>> Objet: Re: [robinhood-support] Issue with accestime criteria
>>
>> On 03/13/14 10:04, [email protected] wrote:
>> > Indeed, it's seems to work. It's quite surprising because in the log
>> > there's still the "wrong" last access, but the policy is apply on the
>> > "good" one.
>> An extract of your policy definition would help me understanding what
>> you want to do,
>> and a copy the related line of log too, so I can double check and ensure
>> your policy application is safe.
>>
>> I've never heard about running robinhood on GPFS. People use to
>> implement built-in GPFS policy engine on it.
>> I'd be curious to know your interests in running robinhood on GPFS, and
>> your feedback about this experimentation.
>>
>> Thanks
>> Thomas
------------------------------------------------------------------------------
Learn Graph Databases - Download FREE O'Reilly Book
"Graph Databases" is the definitive new guide to graph databases and their
applications. Written by three acclaimed leaders in the field,
this first edition is now available. Download your free book today!
http://p.sf.net/sfu/13534_NeoTech
_______________________________________________
robinhood-support mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/robinhood-support
