gLite open issues after EGEE

gLite 3.1 Support after EGEE

gLite 3.2 status

The following table (created on 10.06.2010, it may be obsolete now) shows when a certain gLite service has been released in 3.2. If it has been more than 2 months available in production, I consider it is a candidate to stop its support in 3.1.

(*) Services available in 3.2 but for less than 2 months.

Feedback from EGEE Operations

EGEE Operations (SA1) made a survey before the end of the EGEE project and collected feedback from the sites to understand the feasibility of stopping the support for gLite 3.1. The collected feedback is summarised below.

General feedback from ROC managers

• Many sites are not ready to phase out SL4 WNs, this is HW related. Support might be needed till the end of the year. This includes glite-WN and glite-TORQUE
• A quick check shows 238 SL4 subclusters vs. 386 for SL5

From UKI

• glite-BDII: Our experience with the glite 3.2 BDII running as a top bdii is that it is not as good as the 3.1 version - it suffers from frequent performance degradation and needs restarted to fix. On restart it takes sometime for the BDII to be able to respond to queries correctly as it has to regather the information from all the site bdiis. Requirement to have a similar performance to the gLite 3.1 one before this one stops being supported.

From CERN

• glite-UI: running both gLite 3.1 and 3.2, both being used, but no strong requirement to keep support for gLite 3.1
• glite-WN: still some capacity provided through SLC4, could be migrated in a few months
• LCG-CE: only available in SL4 and still required by some VOs (e.g. Atlas)

From Italy

gLite components for which the request that support in 3.1 continues since many medium and small sites have difficulties in procuring new hardware for the migration of glite 3.2:

• glite-BDII
• glite-CREAM: all Italian CREAM instances are still sl4 but support for gLite 3.1/sl4 could be stopped when CREAM CE 1.7 is released since they will try to be in SL5 by then.
• glite-SE_dpm_mysql: all ATLAS T2 sites currently deploy DPM sl4
• glite-WN: 70% of WNs in Italy are gLite 3.2. The remaining 30% of the logical CPUs are still gLite 3.1, these are mainly deployed in small sites that have difficulties with replacing the existing hw. For this reason we request that the support continues. We don't want to loose the small sites, which are the ones from which a future extension of the infrastructure is expected
• glite-TORQUE_client: support should continue as explained above
• glite-TORQUE_server: support should continue as explained above
• glite-TORQUE_utils: support should continue as explained above
• glite-UI

gLite components for which support can stop:

• glite-GLEXEC_wn: modest impact on production infrastructure, most LHC sites are migrating to sl5 64bit WNs
• glite_LFC_mysql and glite_LFC_oracle: all LFC services operated in the Italian region are still on 32bit sl4 servers, but a migration to 64bit is already planned. If support stops, the migration needs to be rescheduled as soon as possible
• glite-SCAS: it is not deployed
• gLite-VOBOX: we understand from ALICE that all ALICE VOBOX are already sl5. ATLAS is not using the VOBOX outside CERN. The local CMS contacts explained that the sl5 version should work fine. Generally speaking all VOBOX components installed at the INFN Tier1 will be migrated to sl5 in the next couple of months, so the stop of support is not a problem for them. A final work on this for the WLCG computing coordinators is needed.

Summary of services to be supported in 3.1

After combining the information gathered in the two previous sections: 3.2 status + feedback presented by EGEE operations, this would be the final list of services we still need to support in 3.1:

• glite-APEL
• glite-BDII
• glite-CONDOR_utils
• glite-CREAM
• glite-FTS/FTA/FTM
• glite_LSF_utils
• glite-PX
• glite-SGE_utils
• glite-SE_dpm_disk
• glite-SE_dpm_mysql
• glite-TORQUE_client
• glite-TORQUE_server
• glite-TORQUE_utils
• glite-UI
• glite-VOMS_mysql
• glite-VOMS_oracle
• glite-WMS
• glite-WN
• lcg-CE

Feedback from WLCG

WLCG has analisyed the feedback presented in the previous sections and has also studied the needs within WLCG. This is the feedback they have provided:

Service Name	Released on gLite 3.2	Candidate to stop 3.1?	Comments
glite-APEL	07.06.10	NO
glite-ARGUS	08.02.10	Not in 3.1
glite-BDII	29.06.09	YES	Open issues affecting 3.2 BDII should be immediately fixed by developers if support for 3.1 is going to be dropped. GGUS:57870 and GGUS:58951
glite-CONDOR_utils	NO	NO	This was with PIC in EGEE. They say they will do a release for 3.2. No news.
glite-CREAM	07.01.10	YES
glite-FTS/FTA/FTM	21.07.2010	NO
glite-GLEXEC_wn	07.01.10	YES
glite-LB	15.04.10	YES	But major bugs may be needed to be fixed in 3.1 if related with 3.1 WMS.
glite_LFC_mysql	27.07.09	YES
glite_LFC_oracle	08.02.10	YES
glite_LSF_utils	24.03.10	NO	Needed in 3.1 for lcg-CE
glite-PX	NO	NO	This is with Helsinki. They say they will do the release for 3.2. No date yet.
glite-SCAS	07.01.10	YES
glite-SGE_utils	03.05.10	NO	Needed in 3.1 for lcg-CE
glite-SE_dpm_disk	08.02.10	NO	Migration will take a long time so support in 3.1 is needed until sites move to SL5.
glite-SE_dpm_mysql	27.07.09	NO	Migration will take a long time so support in 3.1 is needed until sites move to SL5.
glite-TORQUE_client	12.03.09	YES
glite-TORQUE_server	07.01.09	YES
glite-TORQUE_utils	07.01.09	NO	Needed in 3.1 for lcg-CE
glite-UI	15.06.09	NO	SL5/32bit clients are going to be provided for WMS clients, needed by ATLAS.
glite-VOBOX	13.10.09	YES
glite-VOMS_mysql	03.05.10	NO
glite-VOMS_oracle	18.08.10	NO
glite-WMS	NO	NO	SL5 is in progress
glite-WN	12.03.09	YES
lcg-CE	NO	NO	Do we need to support batch systems in 3.1? Because lcg-CE only in 3.1. (Right now 435 lcg-CEs vs 105 CREAM CEs)

gLite after EGEE

Status of the gLite services after the end of EGEE. We have to make sure all the services not supported in EMI and of interest for WLCG are covered.

Component	Responsible	Status
APEL	RAL	OK
Argus	Switch	OK
BDII	CERN	OK
CONDOR	-	Not present in EMI. Best effort by PIC.
cream CE	INFN	OK
dCache	DESY	OK
DPM	CERN	OK
FTA	CERN	OK
FTM	CERN	OK
FTS	CERN	OK
glexec	Nikhef	OK
LB	CESNET	OK
lcg CE	CERN	OK
LFC	CERN	OK
LSF	-	Not present in EMI. It used to be CERN + INFN in EGEE. INFN is OK to work on this as before. CERN (IT-PES) is studying how to provide their part.
MyProxy	HIP	It will be released in 3.2
SCAS	Nikhef	OK
SGE	-	Not present in EMI. Best effort by LIP.
TORQUE	-	Not present in EMI. Best effort by Nikhef.
UI,WN,VOBOX	CERN	OK
VDT	-	IGE project but not clear what they will provide. In transition period, Alain Roy agreed to help us and CERN will do the integration of new packages in ETICS
VOMS	INFN	OK
WMS	INFN	OK
lcg-tags/lcg-ManageVOTags/lcg-infosites	CERN	OK
gridsite	CESNET	With CESNET in EMI. CESNET was requested on EMT 19.05.2010 to start taking this over from Andrew McNab
proxyrenewal	CESNET	OK
trustmanager	HIP	OK

Are they important for LCG?

-- MariaALANDESPRADILLO - 21-May-2010