Print

Print


Ah, if that feature always works for any redirection response, I will just use that. 

Thanks Andy,
Andreas.


On Tue, Feb 11, 2020 at 1:04 PM Andrew Hanushevsky <[log in to unmask]> wrote:
This seems to be an EOS issue. The object in question does have an internl
2K limit. That was, however, extended to have no limit if one supplies an
external buffer. It would seem that EOS is not using that feature.

On Tue, 11 Feb 2020, Sang Un Ahn wrote:

> Dear Exports,
>
> We are posting this to both of EOS and XRootD supports because it seems to us this issue relates to both.
>
> We found that the number of stripes exceeding 12 when we configure RAIN(12+4) on EOS instance by setting "sys.forced.nstripes=16" gives the following error:
>
> 200207 02:51:22 time=1581043882.201812 func=Emsg level=ERROR logid=b93a52f8-4954-11ea-b68a-b8599fa51330 unit=[log in to unmask]<mailto:unit=[log in to unmask]>:1094 tid=00007ff409bfe700 source=XrdMgmOfsFile:2790 tident=root.140:[log in to unmask]<mailto:[log in to unmask]> sec=sss uid=2 gid=2 name=daemon geo="" Unable to open file - capability exceeds 2kb limit /eos/gsdc/testrain/file1g-jbod-mgmt-09; Cannot allocate memory
>
> And this is the case of success with having "sys.forced.nstrips=12":
>
> 200207 07:56:14 time=1581062174.302108 func=ProcessCapOpaque level=INFO logid=5021a2e6-497f-11ea-9a1d-b8599fa51330 unit=[log in to unmask]<mailto:unit=[log in to unmask]>:1096 tid=00007fce48129700 source=XrdFstOfsFile:2745 tident=4.6:[log in to unmask]<mailto:[log in to unmask]> sec=(null) uid=99 gid=99 name=(null) geo="" capability=&mgm.access=create&mgm.ruid=2&mgm.rgid=2&mgm.uid=2&mgm.gid=2&mgm.path=/eos/gsdc/testrain/file1g-jbod-mgmt-09&mgm.manager=jbod-mgmt-01.eoscluster.sdfarm.kr:1094<http://jbod-mgmt-01.eoscluster.sdfarm.kr:1094>&mgm.fid=00000036&mgm.cid=11&mgm.sec=sss|daemon|jbod-mgmt-09.eoscluster.sdfarm.kr<http://jbod-mgmt-09.eoscluster.sdfarm.kr>||daemon|||eoscp&mgm.lid=1080298322&mgm.bookingsize=1073741824&mgm.targetsize=1073741824&mgm.fsid=1075&mgm.url0=root://jbod-mgmt-09.eoscluster.sdfarm.kr:1096//&mgm.fsid0=1495&mgm.url1=root://jbod-mgmt-01.eoscluster.sdfarm.kr:1095//&mgm.fsid1=67&mgm.url2=root://jbod-mg
 mt-08.eoscluster.sdfarm.kr:1095//&mgm.fsid2=1243&mgm.url3=root://jbod-mgmt-05.eoscluster.sdfarm.kr:1096//&mgm.fsid3=823&mgm.url4=root://jbod-mgmt-06.eoscluster.sdfarm.kr:1096//&mgm.fsid4=991&mgm.url5=root://jbod-mgmt-03.eoscluster.sdfarm.kr:1096//&mgm.fsid5=487&mgm.url6=root://jbod-mgmt-07.eoscluster.sdfarm.kr:1095//&mgm.fsid6=1075&mgm.url7=root://jbod-mgmt-05.eoscluster.sdfarm.kr:1095//&mgm.fsid7=739&mgm.url8=root://jbod-mgmt-02.eoscluster.sdfarm.kr:1095//&mgm.fsid8=235&mgm.url9=root://jbod-mgmt-08.eoscluster.sdfarm.kr:1096//&mgm.fsid9=1327&mgm.url10=root://jbod-mgmt-04.eoscluster.sdfarm.kr:1096//&mgm.fsid10=655&mgm.url11=root://jbod-mgmt-02.eoscluster.sdfarm.kr:1096//&mgm.fsid11=319&cap.valid=1581065774
>
> The error comes from here: https://github.com/cern-eos/eos/blob/master/mgm/XrdMgmOfsFile.cc#L2585
> And XrdOucEI::Max_Error_Len has static size of 2Kb as described here: https://xrootd.slac.stanford.edu/doc/doxygen/current/html/structXrdOucEI.html
>
> FYI, we currently have 18 FSTs out of 9 servers (2 FSTs on each server) and intent to run them with 12+4 (16 stripes) RAIN configuration of EOS for data archive purpose. In near future, we will add 2 more FSTs in order to have a complete setup of 14+4 (18 stripes) out of 20 FSTs.
>
> We were also thinking about having the name of URL (including hostname and domain name) short however we doubt that this resolves the origin of issue. In some cases, the length of filename including its path might be pretty long in production, e.g. ALICE has typical name of its raw data like "/eos/<site::eos::Instance>/03/60094/00eec022-9340-11e5-a2a6-1f249ae2dbab".
>
> So, we are not pretty sure of whether increasing or unbarring the hard limit of XrdOucEI::Max_Error_Len would make sense but it would be very helpful for us if you give us any solutions on this issue.
>
> Thank you in advance.
>
> Best regards,
> Sang-Un
>
>
>

########################################################################
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