Print

Print


Hi Wei,


We are using versions 4.9.1 and version 4.10.0 (on two different clusters)

I think Andy was referring to the following:

When the dataserver is configured for checksum using the external xrdadler32 and 
a checksum is run (remotely in this case), the output from xrdadler32 returns 
three tokens:

[root@admin ~]# xrdadler32 root://storage-5-18//xrd/srm-test/mover-test.1
063301c4 /xrd/srm-test/mover-test.1 root://storage-5-18//xrd/srm-test/mover-test.1


However, when the dataserver is configured for internal checksums and a checksum 
is performed the output only contains two tokens:

[root@admin ~]# xrdadler32 root://storage-5-18//xrd/srm-test/mover-test.1
063301c4 root://storage-5-18//xrd/srm-test/mover-test.1

It looks like ( to my untrained eyes), that when xrdadle32 is be being called 
from within the xrootd process, it returns the two tokens (chksun and 
localpath).  The daemon is then returning the entire output back to to client 
as: (result URL), where result is the two tokens captured from the xrdadler32 
command.

Patrick


 > On 2/19/20 6:31 PM, Yang, Wei wrote:
> Hi Patrick,
> 
> what version of xrootd client (and server) are you using?
> 
> regards,
> --
> Wei Yang  |  [log in to unmask]  |  650-926-3338 (O)
>   
> 
> ´╗┐On 2/19/20, 5:49 AM, "[log in to unmask] on behalf of Patrick McGuigan" <[log in to unmask] on behalf of [log in to unmask]> wrote:
> 
>      Hi Andy,
>      
>       From the host storing the file:
>      
>      [root@storage-5-18 ~]# xrdadler32 root://localhost//xrd/srm-test/mover-test.1
>      063301c4 /xrd/srm-test/mover-test.1 root://localhost//xrd/srm-test/mover-test.1
>      
>      Patrick
>      
>      On 2/19/20 7:12 AM, Andrew Hanushevsky wrote:
>      > Hi Patrick,
>      >
>      > Uhmmm, how about the output of the xrdadler command run on the server against
>      > the same file the query checksum is beig targeted?
>      >
>      > Andy
>      >
>      > On Wed, 19 Feb 2020, Patrick McGuigan wrote:
>      >
>      >> Hi Andy,
>      >>
>      >> [root@admin ~]# xrdcp -np -f --cksum adler32:print test-transfer
>      >> root://storage-5-18:1094//xrd/srm-test/mover-test.1
>      >> Run: [ERROR] Invalid response
>      >> [root@admin ~]# echo $?
>      >> 53
>      >>
>      >> Attached is the output of the same command with the -d3 switch (the
>      >> transferred file is 4 bytes).
>      >>
>      >> Patrick
>      >>
>      >>
>      >> On 2/19/20 5:58 AM, Andrew Hanushevsky wrote:
>      >>> Hi Patrick,
>      >>>
>      >>> Coud run that command by hand on a failing file and post the output as well
>      >>> as what the return code was.
>      >>>
>      >>> Andy
>      >>>
>      >>>
>      >>> On Wed, 19 Feb 2020, Patrick McGuigan wrote:
>      >>>
>      >>>> Hi Andy,
>      >>>>
>      >>>> It is the command from the xrootd-client rpm.
>      >>>>
>      >>>> Patrick
>      >>>>
>      >>>> On 2/19/20 5:45 AM, Andrew Hanushevsky wrote:
>      >>>>> Goo question Rob. We provide the xrdadler command but not a script. So,t he
>      >>>>> first question should have been is this a script or the actual command.
>      >>>>>
>      >>>>> Andy
>      >>>>> On Wed, 19 Feb 2020, Robert Gardner wrote:
>      >>>>>
>      >>>>>> Who provides said external script?
>      >>>>>>
>      >>>>>>
>      >>>>>>> On Feb 19, 2020, at 5:34 AM, Andrew Hanushevsky
>      >>>>>>> <[log in to unmask]> wrote:
>      >>>>>>>
>      >>>>>>> Hi Patrick,
>      >>>>>>>
>      >>>>>>> In the first case, the config says hat an external script is supposed to
>      >>>>>>> computed the checksum. If the external script produces an incorrect
>      >>>>>>> response, the client will get it and complain. The second case says that
>      >>>>>>> the checksum is to be computed using the internal native mechanism. In
>      >>>>>>> this case, a corect response will always be sent. Hence, it will work. We
>      >>>>>>> reccomend that the internal mechanism be used when it is possible to do
>      >>>>>>> so. That said, ot would be interestin to know what the hiccup is in the
>      >>>>>>> external script.
>      >>>>>>>
>      >>>>>>> Andy
>      >>>>>>>
>      >>>>>>> On Wed, 19 Feb 2020, Patrick McGuigan wrote:
>      >>>>>>>
>      >>>>>>>> Hi,
>      >>>>>>>>
>      >>>>>>>> I noticed something odd recently and I am curios if the issue is a known
>      >>>>>>>> feature, or something that needs to be worked on?
>      >>>>>>>>
>      >>>>>>>> A client process was moving files with xrdcp using:
>      >>>>>>>>
>      >>>>>>>> xrdcp -f -np --cksum adler32:print test-transfer
>      >>>>>>>> root://some-server//some-path
>      >>>>>>>>
>      >>>>>>>> and the command fails with:
>      >>>>>>>>
>      >>>>>>>> Run: [ERROR] Invalid response
>      >>>>>>>>
>      >>>>>>>> Tracking this down leads me to the problem in the data server's
>      >>>>>>>> configuration:
>      >>>>>>>>
>      >>>>>>>> xrootd.chksum adler32 /usr/bin/xrdadler32
>      >>>>>>>>
>      >>>>>>>>
>      >>>>>>>> However, if the checksum is configured as:
>      >>>>>>>>
>      >>>>>>>> xrootd.chksum max 4 adler32
>      >>>>>>>>
>      >>>>>>>> The xrdcp will work correctly.
>      >>>>>>>>
>      >>>>>>>>
>      >>>>>>>> Before anyone asks, yes, checksums work correctly outside of xrdcp:
>      >>>>>>>>
>      >>>>>>>> xrdadler32 root://some-server//some-path
>      >>>>>>>>
>      >>>>>>>> as well as:
>      >>>>>>>>
>      >>>>>>>> xrdfs some-server query checksum /some-path
>      >>>>>>>>
>      >>>>>>>>
>      >>>>>>>> I have seen this occur with versions 4.9.1 and 4.10.0.
>      >>>>>>>>
>      >>>>>>>>
>      >>>>>>>>
>      >>>>>>>> Regards,
>      >>>>>>>>
>      >>>>>>>> Patrick
>      >>>>>>>>
>      >>>>>>>> ########################################################################
>      >>>>>>>> Use REPLY-ALL to reply to list
>      >>>>>>>>
>      >>>>>>>> To unsubscribe from the XROOTD-L list, click the following link:
>      >>>>>>>> https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=XROOTD-L&A=1
>      >>>>>>>>
>      >>>>>>>
>      >>>>>>> ########################################################################
>      >>>>>>> Use REPLY-ALL to reply to list
>      >>>>>>>
>      >>>>>>> To unsubscribe from the XROOTD-L list, click the following link:
>      >>>>>>> https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=XROOTD-L&A=1
>      >>>>>>
>      >>>>>>
>      >>>>
>      >>>> ########################################################################
>      >>>> Use REPLY-ALL to reply to list
>      >>>>
>      >>>> To unsubscribe from the XROOTD-L list, click the following link:
>      >>>> https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=XROOTD-L&A=1
>      >>>>
>      >>
>      >> ########################################################################
>      >> Use REPLY-ALL to reply to list
>      >>
>      >> To unsubscribe from the XROOTD-L list, click the following link:
>      >> https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=XROOTD-L&A=1
>      >>
>      
>      ########################################################################
>      Use REPLY-ALL to reply to list
>      
>      To unsubscribe from the XROOTD-L list, click the following link:
>      https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=XROOTD-L&A=1
>      
> 
> 
> ########################################################################
> Use REPLY-ALL to reply to list
> 
> To unsubscribe from the XROOTD-L list, click the following link:
> https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=XROOTD-L&A=1
> 

########################################################################
Use REPLY-ALL to reply to list

To unsubscribe from the XROOTD-L list, click the following link:
https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=XROOTD-L&A=1