EGI TF 2011 Session: Task Force Repositories and Automation ----------------------------------------------------------- Wednesday 21 September 2011 Participants ------------ Michel Drescher, EGI (MD) Kostas Koumantaros, EGI (KK) Marios Chatziangelou, EGI (MC) Enol Fernandez del Castillo, EGI (EF) Matias Ellert, IGE (ME) Helmut Heller, IGE (HH) Stephen Crouch, IGE (SC) Christina Aftimei, EMI (CA) Allessandro Paolini, EMI (AP) Anders Wäänänen, EMI (AW) Agenda ------ - Review the Provisioning process for UMD 1.0, UMD 1.1 and UMD 1.2 - Next steps & Feedback from EMI (skipped) - Next steps & Feedback from IGE (skipped) - Next steps & Feedback from SAGA (skipped) Review of the provisioning process for the UMD ---------------------------------------------- KK gives an overview of the provisioning of UMD 1.0.0, UMD 1.1.0 and UMD 1.2.0 including a summary of issues that were encountered until now and how they were solved. Then MD and KK open the discussion by reflecting the relationship of repositories of EGI, EMI and IGE, and other Technology Providers that may come into the picture at a later point in time. By analysing the process of software published in various repositories the participants conclude with the following: * The EMI and IGE repositories are considered upstream to the EPEL repository (with the EPEL repository being a pre-stage to the SL5/RHEL main distributions) * At the same time the EMI and IGE repositories are also upstream repositories to the UMD repository. * Given the recommended repository configurations for EMI, IGE and UMD being the same the UMD repository must be considered upstream to EPEL as well. The conclusion is that we must ensure that software in EM, IGE and UMD repositories must be of the same version or more recent than the same software present in EPEL. KK and MD propose the policy: "Products in the UMD repository must be at all times of the same or more recent version than the same software being present in the EPEL repository." MD states that this implies that a certain level of coordination is required, so that Technology Providers publish software to EPEL only after it got published in the UMD repository. KK asks why the same RPMs are published in UMD and EPEL. MD replies that EGI has established SLAs with IGE and EMI for provisioning software by these projects as opposed to the maintainers of EPEL. Therefore IGE and EMI are considered the trusted sources for software for the UMD. Current status UMD vs. EPEL --------------------------- MD relays the message from Alberto diMeglio originating from the TCB mailing list that the IGE components currently in UMD are older compared to the ones at EPEL. He further expresses the need for a clear statement whether UMD 1.2.0 as it is is safe for production use, of whether an emergency update is necessary. ME replies that this was caused by IGE not wanting to cause trouble in the EGI provisioning process since Globus 5.03 was already taken in while Globus 5.0.4 was already available in the IGE repositories. MR states that the EGI provisioning process can accept more that one version of a given product in the process in parallel with the only exception that StagedRollout must test one version at a time (which is by design). AP states that EMI Products were certified against IGE 5.0.4 that is now in EPEL. Participants narrowed down the list of the offending packages to the following. UMD EPEL globus-ftp-control-2.11-3.el5.x86_64 globus-ftp-control-2.12-2.el5.x86_64 globus-gass-copy-5.10-2.el5.x86_64 globus-gass-copy-5.14-1.el5.x86_64 globus-gridftp-server-3.28-3.1.el5.x86_64 globus-gridftp-server-3.33-1.el5.x86_64 globus-gridftp-server-control-0.45-2.el5.x86_64 globus-gridftp-server-control-0.46-1.el5.x86_64 globus-gridftp-server-progs-3.28-3.1.el5.x86_64 globus-gridftp-server-progs-3.33-1.el5.x86_64 globus-gsi-sysconfig-3.1-4.el5.x86_64 globus-gsi-sysconfig-3.2-1.el5.x86_64 globus-gss-assist-5.9-4.el5.x86_64 globus-gss-assist-5.10-1.el5.x86_64 globus-gssapi-error-2.5-7.el5.x86_64 globus-gssapi-error-2.5-8.el5.x86_64 globus-gssapi-gsi-7.6-2.el5.x86_64 globus-gssapi-gsi-7.8-1.el5.x86_64 globus-proxy-utils-3.9-3.el5.x86_64 globus-proxy-utils-3.10-1.el5.x86_64 Based on this list of packages the participants of the meeting conclude that UMD 1.2.0 in its current state is reasonably safe to use with no expectation of any issue arising in the production infrastructure from that circumstance. Further, EGI and IGE agree that Globus 5.0.4 will be provisioned as soon as possible to be included in UMD 1.3.0 (scheduled for 31 October 2011), or if necesary, to be published as an immediate update as required. Any other business ------------------ All other planned agenda items were skipped in favor of finding a conclusion for the state of UMD 1.2.0.