Print

Print


Hi Dan,

Dan Schrager wrote:

>
> Let's say that I would like to get aligned to the new default, i.e. I 
> want to kill gridice on $SE_HOST.
> I repeat my questions remained unanswered for now:
> 1) which RPMs have to be erased from the $SE_HOST, exclusively related 
> to this gridice service;

you don't have to erase any rpm, the two packages:

edg-fabricMonitoring-2.5.4-4
gridice-sensor-1.5.1-pl2_sl3

are installed both on client and collector nodes.
On client nodes (the ones that you want to monitor) you have 
edg-fmon-agent running.
On the collector node you have 2 more proceses:
edg-fmon-server and gridice-mds.

If you want to move the collector you have to stop this 2 processes, and 
chkconfig --del <them>. A new configure_node will take care of some 
configuration files but won't stop this 2 processes.

> 2) does YAIM remove RPMs no longer needed in the process of 
> configure/install of a node ? (and which might stay in the way of an 
> upgrade or a simple move of a service from one node to another -- like 
> this gridice)

No.
YAIM is using apt-get to install a node, it will abort in case of 
conflicts between packages giving suggestions, but will not remove 
packages. Until now gridice packages didn't have any conflicts.

> 3) does YAIM at least chkconfig --del <unwanted_services> ?
>
No - unfortunatelly you have to do as you say below.

> I would do such changes by hand, thank you very much, :-)   , provided 
> that somebody answers (1)...
>
>
All the Best,
Cristina

PS: in case of more problems with gridice - there is [log in to unmask]

> Sergio Andreozzi wrote:
>
>> Hi Laurence,
>>
>> Laurence Field wrote:
>>
>>> The reason for this decision was to make things more uniform.
>>>
>>> MON in this case is short for Monitoring.  Previous to LCG-2_6_0, 
>>> this only contained the R-GMA server. However, it would makes sense 
>>> to move the GridIce server here too and reduce the load on the SE.  
>>> This will give us a more uniform structure.
>>> CE is the interface to the batch system.
>>> SE is the interface to the storage system.
>>> MON is the interface to the monitoring system.
>>
>>
>>
>> just to double-check, please verify that there is still a "standard" 
>> GRIS (CE/SE) publishing the existence of the GridICE collector (alias 
>> the GRIS started by gridice-mds) by using the GlueService entry.
>>
>> Until now, the two GRIS's were running on the same machine. I don't 
>> remember if there is any dependency on the configuration process.
>>
>> Cheers, Sergio
>>
>>>
>>> The default Yaim configuration now does this, however as usual it is 
>>> up to sites to choose.  Some sites will run all services on one 
>>> node,  others run them on separates nodes.   An example is the site 
>>> BDII.  By default this is installed on the CE but some sites chose 
>>> to install it on a separate node.
>>>
>>>
>>> Laurence
>>>
>>>
>>> Dan Schrager wrote:
>>>
>>>> Hi YAIM authors,
>>>>
>>>> I have noticed that yaim 2.4.0 sets by default 
>>>> GRIDICE_SERVER_HOST=$SE_HOST while the latest yaim 2.6.0 sets it to 
>>>> GRIDICE_SERVER_HOST=$MON_HOST (in file site-info.def).
>>>>
>>>> Does it matter ?
>>>> Should I move the gridice service to the $MON_HOST and terminate it 
>>>> on the $SE_HOST ? Which rpms should be erased from the $SE_HOST, 
>>>> related exclusively to this service ? Does YAIM automagically do 
>>>> this work during install and configure (at least it stops 
>>>> permanently the related services ?)
>>>> Should I leave it on the $SE_HOST ?
>>>>
>>>> Regards,
>>>> Dan
>>>>
>>
>>
>


-- 
---
Cristina Aiftimiei - EGEE Project
Ist. Naz. di Fisica Nucleare - Padova
Address:    via F. Marzolo, 8 - 35131 Padova - ITALY  
Email: [log in to unmask]
Phone: +39-0498277005
Mobile:+39-3460230488