italiangrid / puppet-creamce Goto Github PK

Using puppet 5 I've got the following error

Error: Could not retrieve catalog from remote server: Error 500 on SERVER: Server Error: Evaluation Error: Error while evaluating a Function Call, Could not autoload puppet/parser/functions/build_group_definitions: no such file to load -- gridutils (file: /etc/puppetlabs/code/environments/prebatch/modules/creamce/manifests/poolaccount.pp, line: 60, column: 18)

I managed to fix it just adding the ruby module path to the ruby path env on the puppetserver

[root@puppetmaster-2 environments]# grep ruby-load-path /etc/puppetlabs/puppetserver/conf.d/puppetserver.conf
ruby-load-path: [/opt/puppetlabs/puppet/lib/ruby/vendor_ruby, /etc/puppetlabs/code/environments/prebatch/modules/creamce/lib/]
[root@puppetmaster-2 environments]#

It must be considered a workaround.

Ale

Dear All

Currently number of cores published by CREAM node is determined from
nodes list in creamce::hardware_table and ce_physcpu + ce_logcpu

Would it be possible to create override variable to handle this?
In our case when we haven't specified nodes in the hardware_table value is presented as 0

[root@ce01 ~]# ldapsearch -x -H ldap://127.0.0.1:2170 -b o=grid | grep HEP-SPEC
GlueHostProcessorOtherDescription: Cores=0,Benchmark=8.07-HEP-SPEC06

creamce::site::name defaults to fqdn when not set what may cause issues when going production.
When that happens site is visible in bdii but some tools (i.e for atlas) are not able to discover such cream endpoint properly and assign it to a site

Please consider printing out warning instead of setting default value

On some installations where poolaccount home directories are located on shared storage (NFS/Lustre/etc.) there is no need to use scp nor any other stagein/stageout mechnism.

Example:
In our case we have slurm + nfs home dirs so we need to be able to define the following variable in /etc/blah.conf to prevent job script from calling scp:

blah_shared_directories="/home/grid"

Please consider adding blah::shared_directories configuration variable to address this

Using puppet 5 I've got two "Unknown resource type". In particular

• Unknown resource type: 'pooluser'
• Unknown resource type: 'vofiles'

I managed to fix it by creating a file per each define type

[italiano@ui01 prebatch]$ head -1 modules/creamce/manifests/vofiles.pp
define creamce::vofiles ($server, $port, $dn, $ca_dn, $gtversion, $voname, $vodir) {

[italiano@ui01 prebatch]$ head -1 modules/creamce/manifests/pooluser.pp
define creamce::pooluser ($uid, $groups, $gridmapdir, $comment,
[italiano@ui01 prebatch]$

it is the recommended way in puppet 5

Ale

Yaim configuration generator allowed to define custom prefix for the names of job submitted to a cluster from particular cream node.

This is important when site is deploying more than one cream node on the same batch system.

For example:
In the yaim config file we defined the following variable
BLAH_JOBID_PREFIX=crm02_

Configuration generator in case of PBS created the following entries in configuration files:

/etc/lrms/scheduler.conf:cream_prefix: crm02_
/etc/lrms/pbs.conf:cream_prefix: crm02_
/etc/glite-ce-cream/cream-config.xml:      <parameter name="BLAH_JOBID_PREFIX" value="crm02_" />

Currenty in case of slurm we need to manually change parameters in scheduler.conf and cream-config.xml after each pupper run

Please consider adding blah::jobid_prefix variable to address this requirement.

Would you please add a parameter to define grid user home dir?
Thanks very much!

Xiaofei