Hola Juanjo,
>>> I have installed a cream-ce. I can see that the atlas users, who are
>>> mapped to atls or atlxxx are OK, but with a user with production role it
>>> is not OK.
>>>
>>> [...]
>>>
>>>> uberftp grid001.ft.uam.es
>>> 220 grid001.ft.uam.es GridFTP Server 2.8
>>> (gcc64dbg, 1217607445-63) [VDT patched 4.0.8] ready.
>>> 530 Login incorrect. : Unable to set the gid of the server process.
>> I suspect it is due to the contents of /etc/group on your CE:
>> it has a huge number of entries for the "atlas" group, in particular
>> entries that are not needed, viz. the pool accounts for ordinary users!
>>
>> A few years ago we ran into a similar problem with YAIM (check comment #3):
>>
>> https://savannah.cern.ch/bugs/index.php?27056
>>
>> As a test you can remove all the ordinary pool accounts from the atlas group
>> and do the same for dteam; then check if atlas production jobs succeed.
>
> Try this first:
>
> perl -e 'use POSIX; print sysconf("_SC_GETGR_R_SIZE_MAX") . "\n"'
That turns out not to work. Try this C program instead:
----------------------------------------------------------------------
#include <stdio.h>
#include <unistd.h>
main()
{
return !printf("%d\n", sysconf(_SC_GETGR_R_SIZE_MAX));
}
----------------------------------------------------------------------
I get 1024 on our CEs...
|