Node name Post-migration cleanup

[nataliaratnikova]

PhEDEx hard-codes t_adm_node.se_name value during node creation, and updates can only be done manually by the global admin. After Node name transitioning this field should no longer be used or exposed to the users.

Post-migration cleanup: Remove SE information from DB, APIs, agents, everything…

[nataliaratnikova]

Tested in testbed datasvc installation.

[nataliaratnikova]

Closed by accident, reopening.

[nataliaratnikova]

List of places where the old se_name from t_adm_nodes table is currently used: 1) Utilities/NodeNew - need to remove --se-name option, update usage help. Looking at the code, utility is using oracle procedure, introduced in PHEDEX_4_1_1 according to Documentation/ChangeLog. Do we need to update the procedure as well? Here it goes:

SQL> describe PROC_ADD_NODE;
PROCEDURE PROC_ADD_NODE
 Argument Name          Type            In/Out Default?
 ------------------------------ ----------------------- ------ --------
 NAME               VARCHAR2        IN
 KIND               VARCHAR2        IN
 TECHNOLOGY         VARCHAR2        IN
 SE_NAME            VARCHAR2        IN

SQL> describe PROC_DELETE_NODE;
PROCEDURE PROC_DELETE_NODE
 Argument Name          Type            In/Out Default?
 ------------------------------ ----------------------- ------ --------
 NODE               VARCHAR2        IN

SQL>

For code defining the procedures look in https://github.com/dmwm/PHEDEX/blob/master/Schema/OracleCoreFunctions.sql#L39-L62

[nataliaratnikova]

2. Schema: there is a V_DBS_BLOCK_REPLICA view that contains se_name, see https://github.com/dmwm/PHEDEX/blob/master/Schema/OracleCoreViews.sql . Is it still needed/used?

SQL> select * from all_objects where object_type='VIEW' and object_name='V_DBS_BLOCK_REPLICA';

OWNER                  OBJECT_NAME            SUBOBJECT_NAME              OBJECT_ID DATA_OBJECT_ID OBJECT_TYPE     CREATED   LAST_DDL_ TIMESTAMP       STATUS  T G S  NAMESPACE EDITION_NAME
----------------

CMS_TRANSFERMGMT           V_DBS_BLOCK_REPLICA                       310918            VIEW        26-SEP-08 19-JAN-11 2010-09-28:14:26:45 VALID   N N N      1

SQL> describe V_DBS_BLOCK_REPLICA;
 Name                                                            Null?    Type
 --------------------------------------
BLOCK_NAME                                   NOT NULL VARCHAR2(1000)
DATASET_NAME                                     NOT NULL VARCHAR2(1000)
SITE_NAME                                    VARCHAR2(4000)
SE_NAME                                  VARCHAR2(80)
SQL> 

SE_NAME can be null.

[nataliaratnikova]

3. StorageConsistencyCheck script - remove --se-name option, update usage help. All checks can be done based on node name.

[nataliaratnikova]

4. Data service APIs:

   • RequestActions - remove from output in the attributes section
   • GroupUsage - remove both from input and output
   • NodeUsage - note about adding se_name to the output - no need any more (node ID may be still useful) https://github.com/dmwm/PHEDEX/blob/master/perl_lib/PHEDEX/Web/API/NodeUsage.pm#L69
   • FileReplicas, BlockReplicas, MissingFiles, RequestList

5. Data Service queries:

   • se_name is all over the place in https://github.com/dmwm/PHEDEX/blob/master/perl_lib/PHEDEX/Web/SQL.pm , perhaps could be left alone? To be checked. getSiteReplicas and getNoes functions in https://github.com/dmwm/PHEDEX/blob/master/perl_lib/PHEDEX/Core/SQL.pm

6. Web site : couple occurrences in https://github.com/dmwm/PHEDEX/blob/master/Documentation/WebSite/PlotConfig/src/phedexgraphs/web/__init__.py https://github.com/dmwm/PHEDEX/blob/master/Documentation/WebSite/access25