Hi Emyr,
> ... so as not to interfere with our LDAP setup.
Just in addition to what previous people have said - we have our accounts in
our LDAP (used to be in NIS) but we migrated.
No real issues, but we had to modify the YAIM config_users to create the
homedirs if the user existed but the dir didn't.
Yours,
Chris.
> -----Original Message-----
> From: Testbed Support for GridPP member institutes [mailto:TB-
> [log in to unmask]] On Behalf Of emyr.james
> Sent: 01 November 2011 14:20
> To: [log in to unmask]
> Subject: Install : User Setup
>
> Hi,
> I'm setting up the cream-ce and need to set up the users.
> CONFIG_USERS=yes doesn't work so am going to set them up manually. I
> intend to generate 2 files I can simply append to /etc/passwd and
> /etc/groups with the right information in, and all the UID's and GID's
> being >70k so as not to interfere with our LDAP setup.
>
> I have stuff like the following in a users.conf file that Chris Walker
> handily let me have a look at...
>
>
> 21001:atlassgm:21001:atlas:atlas:sgm:
> 21002:atlas001:21001:atlas:atlas::
> 21003:atlas002:21001:atlas:atlas::
> .
> .
> 21500:atlas499:21001:atlas:atlas::
> 21501:atlas500:21001:atlas:atlas::
> 21702:prdatl01:21701,21001:prdatl,atlas:atlas:prd:
> 21703:prdatl02:21701,21001:prdatl,atlas:atlas:prd:
> .
> .
> 21750:prdatl49:21701,21001:prdatl,atlas:atlas:prd:
> 21751:prdatl50:21701,21001:prdatl,atlas:atlas:prd:
> 21802:pilatl01:21801,21001:pilatl,atlas:atlas:pilot:
> 21803:pilatl02:21801,21001:pilatl,atlas:atlas:pilot:
>
> etc. etc. with similar blocks for the other VO's (I'm configuring for
> dteam, ops, atlas and snoplus). Obviously I'll be changing the UID's
> and GID's...
>
> I have a few questions....
>
> Do these users need home directories setup (my guess is yes) ?
> Do they all need separate home directories or can they share a
> directory e.g. set homedir for all the atlas users to /home/atlas say
> as opposed to needing /home/atlas001, /home/atlas002 and so on ?
> Should there be one instance of home directories shared by all the
> nodes via nfs/lustre (I'm guessing this is the case...) as opposed to
> local ones on each compute node ?
> What should the login shell be set to ?
>
> Thanks,
> Emyr
|