Minutes Globus integration task force 2011-05-04 10:00-12:00 Minutes belonging to Agenda under https://www.egi.eu/indico/conferenceDisplay.py?confId=469 Presence: Michaela Barth MB Gert Svensson GS Helmut Heller (IGE, LRZ) HH Emmanouil Paisios (IGE, LRZ) EP Ioan Muntean (IGE, UTCN) IM Alex Stanciu (ICI, Romania) AS Marcin Radecki MR Foued Jrad FJ Emir Imamagic EI Torsten Antoni TA Phonebridge: NIKHEF: Oscar Koeroo OK and Mischa Sallé MS Tiziana Ferrari TF 1) Welcome and round the table First meeting of the Globus integration task force as proposed in the last OMB https://www.egi.eu/indico/conferenceDisplay.py?confId=267. Goal: to drive forward Globus integration in EGI. Volunteering members were Croatia, Germany, Romania, Poland, Netherlands, UK. The IGE project will participate to the works of the task force with one member. At the OMB T. Ferrari presented information on Resource Centres that are currently deploying Globus, and about Centres that may be potentially interested in deploying Globus – see RCs currently deploying Globus: https://www.egi.eu/indico/getFile.py/access?contribId=4&resId=0&materialId=1&confId=267 RCs interested in Globus: https://www.egi.eu/indico/getFile.py/access?contribId=4&resId=0&materialId=0&confId=267 Action (all NGIs in the OMB): All NGIs are kindly requested to get in contact with those Centres to understand their interest in being integrated into the NGI infrastructure, and through the NGI, into EGI. MB: The information source given for this list was H. Heller, IGE Project. Helmut was this from a recent survey? HH: This data was presented in the Technical Coordination Board, but stems from the initial EGI survey 3 years ago. The lists contain not not only EGI sites, but also sites from the HPC area, meaning sites connected to DEISA, PRACE and so on, they are not necessarily part of EGI. Round the table: MB: EGI.eu Coordination of Interoperations between NGIs and with other DCIs EP: LRZ in Helmuts Group, administrationg Globus within D-Grid IM: Here for Romania, active in IGE and responsible for computing projects in DEISA, where we use GLobus as well, leading a group of application developers for Globus. Users of our applications expressed their wishes to use Globus resources in Romania. HH: Leader of IGE project, also from LRZ, which is one of the resource center sites in NGI-DE, interested to get the Globus toolkit into UMD and EGI. GS: Previous experience as Northern European ROC manager during EGEE area, deputy director of PDC, will replace Michaela starting this summer, when she goes on maternal leave for some months. AS: from Bucharest, Romania. There is interest from Romania regarding Globus integration task force. 10:11:18] Gert Svensson the sounds is really bad [10:11:24] Helmut Heller it is not understandable [10:11:32] Alex Stanciu ok [10:11:34] Helmut Heller i think your mike gain is too high [10:11:39] Alex Stanciu I am working in ICI Bucharest, Romania [10:11:51] Alex Stanciu involved in operations, as NGI_RO operations manager MR: representing the Polish NGI. One Polish site is involved in the IGE project, this site is taking part within IGE as a testbed organisator, but there exist no plans to provide production resources to the end user. We don't have Globus users in Poland on board. (Which is a different situation compared to UNICORE where we have a lot of production resources and users.) FJ: Working in D-Grid project, responsible of administration of our Globus resources, we provide Globus for D-Grid sites and users, we don't have Globus 5 yet, but plan to integrate it in the next months. I also work for NGI-DE operation, and especially monitoring. EI: probably wearing 3 hats: first as represtantative for Croatia. Croatia has more than 5 years experience with Globus, we have national resources that are just using Globus (GLobus 4), but we are planning to use Globus 5 second hat is monitoring: responsible for integrating new probes into SAM/Nagios third hat being SA1 task leader for monitoring TA: Management of GGUS. GGUS is used as the central helpdesk in EGI and is also basis of the technology helpdesk, it enables communication of bugs to the technology providers and communication of new features and releases from the technology providers. I'm also the project manager for the D-Grid integration project and head of NGI-DE. OK: Oscar and Mischa from NIKHEF, replacing David Groep who is on holiday. Dutch researchers still use GRAM 2 interface (lcg-ce as Globus gatekeeper) to our cluster infrastructure within our NGI, we would still like to support their interface for them or provide at least that functionality that is provided with that and keep supporting it. We would like to offer a seamless transition for our researches to move to a newer version of the middleware stack. IGE could also take a look and adapt our job wrappers, which were created within the WLCG environment. No changes were made with respect to the Globus 4 version of the Gatekeeper. MB: 2) The wiki at https://wiki.egi.eu/wiki/Globus_integration_task_force is thought to keep a list of open actions. I will add you all to the mailinglist globus-integration-tf@mailman.egi.eu 3) Overview status: a) GOCDB https://wiki.egi.eu/wiki/GOCDB/Input_System_User_Documentation#Service_types The three most important service types for Globus which have been registered into the GOCDB are: 1. job submission service for Globus version 4.0.x, 4.2.x (WS-GRAM) and 5.x (GRAM5). 2. storage endpoint and data transfer service for the Globus middleware stack (GridFTP). 3. certificate based interactive login service (gsisshd). Used ports can differ from the default, thus the registration of the port must be possible as well. What should be the next candidates to be defined and supported in GOCDB? Milestone MS407 lists • MDS: Monitoring and Discovery Service for Globus Resources • WebMDS: Web interface to MDS information • RLS: Replica Location Service • RFT: Reliable File Transfer HH: RFT has been superseded by Globus online, MDS will be replaced by IIS, but that is not yet in the status of integration. EI: Helmut, can you give us more details on IIS? HH: IIS is not WS-based any more. We hope to be able to hand out more information at the Community Forum. EI: Are there any rough timelines for it? HH: I don't know by heart --> New AP: HH (or somebody else from IGE) to provide more details and timelines on IIS in the https://rt.egi.eu/rt/Ticket/Display.html?id=390 RT ticket and our mailinglist after the EGCF1 MB: So is it already time and need to register the RLS and Globus Online in GOCDB? HH: We are fine with the current state. I suggest for the moment we start with those which we have. Once we have them fully implemented with monitoring and everything we can add RLS and Globus Online. MB: b) Monitoring: https://wiki.egi.eu/wiki/PROC07_Adding_new_probes_to_SAM Next step is to define tests per service. Here's what we have now according to https://rt.egi.eu/rt/Ticket/Display.html?id=390 GRAM5 * hr.srce.GRAM-Auth * hr.srce.GRAM-Command * hr.srce.GRAM-CertLifetime * hr.srce.GridFTP-Transfer (do GRAMs have GridFTP by default?) globus-GRIDFTP * hr.srce.GridFTP-Transfer globus-GSISSHD * org.nagios.gsissh-Check (check_ssh) EI: I have same questions to the authorization infrastructure, but I could switch these probes on in the next SAM release, which should be released by the end of this month. TF: Emir, can you please summarize open action points and issues at the moment. EI: From monitoring perspective, there is no issue. In the beginning Ilya provided some probes, but we realized we already have those probes, and now we basically can switch them on. TF: Are the tests semantically equivalent to if you compare to gLite and ARC probes, like e.g. for job submission? EI: In Globus you only submit to the Globus gatekeeper, so we submit a test job there. TF: But the job submission task should be valid for different implementations, shouldn't the tests be semantically equivalent? EI: There are two issues: in Globus we have no information system, we can not even tie this to the WMS, we can only tie the probes directly to the service, which from my point of view is a good thing. And there is no lfc-client, so even if we are able to tie this to a WMS, we couldn't use the same SAM/Nagios probes. TF: We want to test a given workflow, like the submission of a job, or the removing of files, can you do this? Are the probes equivalent from a functional point of view? EI: Yes, they are. But you can't map the tests for a probe one to one. TF: But from a high point of view, they are similar? EI: That's correct. TF: Who is maintaining the probes? EI: All this probes were developed by me, this was though most efficient until we see problems and can start offloading to the EGI infrastructure. Part of them could go into the hands of EMI now. Those probes which can't used against EMI services, can be pushed into NGI hands. TF: Helmut, the responsibilities for Nagios probes for functional capabilities are not part of the EMI description, we should clearly identify the responsibility of the efforts. HH: Ilya or Foued from NGI-DE could take over. FJ: No, I will not work for NGI-DE in the future, I have other plans. EI: Helmut, my plan was to first integrate them, then someone from your site can take over. Let's get it in first, then try to offload it elsewhere. If you would just agree to take them over afterwards. HH: I agree with this approach, we are very happy with that. c) GGUS MB: Should have been handled by EMI. Is there someone who tested it, who/which email list gets the email? TA: There is a support unit for IGE, I don't know who is behind the mailinglist. Helmut, do you know? HH: We have had a couple of initial tickets which came over to IGE, it is a mailinglist within IGE. TA: The initial problem with inappropriate tickets is settled, everything goes to DMSU first. Only if it is a bug it is assigned to IGE. MB: Putting down an AP on me to test, just to confirm. --> new AP: MB to test GGUS leaf to IGE: what is the mailinglist and who is on it. d) Accounting MB: The current information I have is still that Globus currently has no accounting software. Accounting statistics for Globus can be obtained indirectly through the underlying batch system. There were efforts of adopting DGAS for Globus in the scope of D-Grid. It was also planned to use OGF-UR there (which was unfortunately not yet provided by DGAS at that time). Nowadays DGAS is part of the EMI project which is aiming to harmonize components to a common standard set. EMI has now in their roadmap agreed on a common usage record based on the OGF-UR standard with extensions. HH: In IGE we are developing the Grid-SAFE, which will use the UR format, but at the moment we don't have anything like that. EP: About Grid-SAFE: We plan installation of it in our testbed, afterwards we can give an update about it in the mailinglist. EI: Who is developing it? HH: Grid-SAFE was developed in the UK, we will integrate it in our IGE release. TF: So that is a complementary function by the UK? HH: Exactly. [10:42:54] Emir Imamagic http://gridsafe.sourceforge.net/about.html ? [10:43:22] Emmanouil Paisios Yes that's the website [10:44:21] Emir Imamagic could someone provide link to packages which IGE is providing? [10:49:17] Emmanouil Paisios Emir, packages for gridsafe are not there yet will be there in the future [10:49:29] Emmanouil Paisios they will be provided by SOTON [10:49:54] Emir Imamagic [10:49:59] Emir Imamagic that's sounds promising [10:50:08] Emir Imamagic the name of package that is [10:50:35] Emmanouil Paisios [10:50:39] Emir Imamagic project I meant, ah too early for me [10:51:33] Emmanouil Paisios SOTON is the University of Southampton, they are participating in IGE (too early for me too ) TF: The problem is that every different integration adds additional fields and extension to the OGF-UR standard. I suggest to avoid a problem with different extensions fore every new application in the future. EI: How do you distribute URs, which transport mechanism do use right now? HH: I can't answer right now to the transport mechanism question. I'll pass it on within IGE. MB: Back to the extensions of the OGF-UR standard. I think I can answer there. John Gordon told me half a year ago that APEL uses the extensions they also suggested to the OGF. So I think I can give you the link [10:45:50] Emir Imamagic https://wiki.egi.eu/wiki/APEL MB: No not that one, one in gridforge. I'll look it up and send it to you Helmut. TF: In one year time from now, we'll still see a context when APEL extensions are not standardized. So what extra information is missing? MB: Extensions are definitely needed. The OGF-UR v 1 doesn't even explain what a site is. For example. So we probably have to push the standardization procedure. TF: This has to be done in parallel. We need action to have this work starting, HH could check the extensions suggested and used by APEL and check what more is needed, from an accounting point of view. HH: As soon as Michaela sends me the link I'll check within IGE if those extensions are sufficient for use in Grid-SAFE. --> new AP: MB to send link with suggested extensions to OGF-UR by APEL to HH. HH to check within IGE if those extensions are sufficient for use in Grid-SAFE. OK: Following up on the question on transport layer that was briefly discussed: based on what I have seen on APEL clients and server installations an non-open standard protocol is used. Do we push any change for messaging technology from ActiveMQ to an open standarized messaging format? Shouldn't we use open standardized formats only? TF: Just one comment here: at the moment the messaging structure is also used by other operational tools (like the dashboard or GGUS) so that is not going to change within one year time. OK: All kinds of software buildings are made in some way. We have to have a long term plan: Would there be a standardized protocol desirable? TF: This work is included in the EMI project, where the clients for accounting are developed, addtionally there is work on messaging going on as well. EGI-Inspire is planning to schedule an Accounting Workshop for the Technical Forum in September in Lyon. This is part of EMI though, and not of EMI Inspire. OK: Wheter we focus on a specific APEL application or focus on an open protocol, that is a decision EGI has to make. TF: I feel we need a dedicated meeting for this. Some of the countries have developed their own regional solution where the UR are collected and then from one site sent collectedly to the APEL central, this could be an interim solution. More work is going on. A new messaging format can be discussed. EI: The messaging itself is just a transport layer, it doesn't define a format. At this point there are solutions were the APEL client is used to publish summaries only. The messaging layer and how to use it is investigated by EMI, I wouldn't say it is not standardized, and I wouldn't say it binds you. EI: It uses STOMP, but the fact that we are using ActiveMQ right now, can be changed. OK: It would be interesting to follow this discussion. EI: also switching to STOMP (Streaming Text Orientated Messaging Protocol) OK: David Groep wrote the UR-push from some site to APEL database directly, this is how we publish their URs. The details of APEL weren't sufficient for our sites. EI: It would be really good to have John or Cristina from APEL here. MB: I suggest to have a next meeting with focus on accounting together with them. e) Early Adoption Group of Globus EA teams: early-adopters-globus@mailman.egi.eu https://www.egi.eu/earlyAdopters/table MB: I can't really say how urgent this is at the moment, but Mario David always insists to have at least two Early Adopters per service. Currently for Globus we have already one. Do we have any more volunteers? TA: I will ask around if someone is interested in doing this. [11:03:00] foued Jrad I suggest to ask uni dortmund [11:03:13] Helmut Heller yes, they are VERY likely to do this. [11:03:55] Tiziana Ferrari yes good [11:04:10] Helmut Heller yes TF: What is the timescale foreseen by the NGIs represented here? When is NGI-DE planning to have Globus, the same question is also valid for Poland, Netherlands, Croatia and Romania. Once we have a timeline we can push staged rollout further. EI: I agree with Tiziana, in order for participating in EA we would need the timelines. AP: ask NGI on timelines, MB: --> new AP: MB to ask NGIs on desired timelines for Globus integration to determine need for a second Early Adopter team for Globus at current stage. --> new subAP: TA to ask within NGI-DE for a second EA volunteer. 4) First Globus community Forum Meeting 18-19th of May, Munich http://www.ige-project.eu/events/egcf1 MB: Helmut, can you give us some more details? HH: This will be a community forum meeting, all Globus users should come or send somebody. There will be a Globus online demonstration and elected speakers. This is definitely interesting for everyone being part of the IGE project and the steering user community. The meeting is held under the umbrella of the yearly NGI-DE meeting. I would like to welcome everybody. There is also the UNICORE forum at the end of Friday which might be of interest to some. MB: It will be in English, right? HH: The Globus part will be in English. TA: For the NGI-DE meeting, English will be spoken if we have non-german speaking particiapants in the workshop on demand. An there is keynote presentation from SN, which will of course be in English. I don't know either for the UNICORE part. HH: If you all could advertise it within your communities. We have a hard time reaching Globus users in Europe. [11:07:33] Tiziana Ferrari I distributed that announcement to all NGI operations managers [11:07:38] Tiziana Ferrari shortly after the TCB [11:08:22] Helmut Heller beer gardens will be open already in munich!! MB: --> new AP on everybody: Advertise First Globus community Forum Meeting http://www.ige-project.eu/events/egcf1 Going on in the Agenda: 5) Next Meeting Topics? So we already have Accounting as hot topic for the next agenda. Anything more? EI: What are the plans for integration in EGI infrastructure; Globus uses Gridmapfile, but a VOMS plugin was developed, how is it planned to do this in the EGI infrastructure? OK: Just as a quick summary: The support for the GT4 and GT5 interface was moved from EMI to the IGE project. Through this interface, we can do all the things we did for the last 6 years, as gLite uses VOMS. So we should take the opportunities to go from there. EI: Will it be part of the first IGE release? OK: Yes. HH: The first IGE release was announced for end of April, but we are still finishing it. This first release will not contain it, but the second one at the end of this year will contain it. EI: This first release will go through the software release procedure and through the OMB, I assume. MB: So, we could have an update on the authorization efforts going on as well. When do we want to have this meeting, shall I make a doodle poll? TF: We could have a dedicated meeting about accounting, otherwise we should follow up the topic about IIS, and what are the implications of having an integrated infrastructure monitoring system and its compatibility with EMI. OK: NIKHEF is involved in EMI. HH: Could we have a summary of todays Actionspoints, so far? (MB tries to give a possibly non complete list of Actionpoints) MB: I'll put them on our wiki page immediatly after the meeting. TF: Could you organize the meeting with the accounting people, what are viable options there, possibly we also have to talk about messaging. HH: We should have some time availible, at the EGCF1 in Munich, we could arrange for that to have a meeting were, were you could phone in. TF: That would be an option. Otherwise I suggest to have it in last week of may before the all hands of EMI F2F. So we already could discuss the outcome there. MB: --> new AP: MB to organize the next meeting as a dedicated accounting meeting with the accounting experts: either as a Phoneconference slot during the EGCF1 or in week 21 before the 2nd EMI allhands F2F in Lund. Just a quick question around the table, who of you would be in Munich? I won't, even if I would really love to fly there. 11:19:16] Helmut Heller me [11:19:17] Ioan Lucian Muntean I'll be there [11:19:24] Emir Imamagic no money, phone conference would be good [11:19:32] Tiziana Ferrari not me [11:19:34] Emmanouil Paisios me too [11:19:38] Gert Svensson Maybe [11:19:41] foued Jrad me tto [11:19:48] Emmanouil Paisios I'm going to munich [11:19:57] foued Jrad im going TF: I will be in Brussels this week, but I would be able to join too. OK: Me and Mischa won't go to Munich, but our expert Denis is going to Munich from NIKHEF. [11:20:29] Tiziana Ferrari I would suggest we ask Gordon and Del Cano first [11:20:29] Alex Stanciu i cannot attend [11:20:33] Tiziana Ferrari as they have to be present [11:20:42] Tiziana Ferrari depending on their availability we schedule the meeting MB: right. 6) AOB [11:21:42] Tiziana Ferrari not from me it was already a good start! MB: Thank you all for participating in this first interesting Globus task force meeting! -------------------------------------------------------------- Open Actionpoints after this first meeting: -------------------------------------------------------------- * AP on everybody: Advertise First Globus community Forum Meeting http://www.ige-project.eu/events/egcf1 * AP: MB to test GGUS leaf to IGE: what is the mailinglist and who is on it. * AP: EP to report on installation of Grid-SAFE in the LRZ testbed to our mailinglist when done. * AP: HH to pass on question to IGE about which transport mechanism is intended to be used in Grid-SAFE. * AP: MB to send link with suggested extensions to OGF-UR by APEL to HH. HH to check within IGE if those extensions are sufficient for use in Grid-SAFE. * AP: HH (or somebody else from IGE) to provide more details and timelines on IIS in the https://rt.egi.eu/rt/Ticket/Display.html?id=390 RT ticket and our mailinglist after the EGCF1 * AP: MB to ask NGIs on desired timelines for Globus integration to determine need for a second Early Adopter team for Globus at current stage. o AP: TA to ask within NGI-DE for a second EA volunteer. * AP: MB to organize the next meeting as a dedicated accounting meeting with the accounting experts: either as a Phoneconference slot during the EGCF1 or in week 21 before the 2nd EMI allhands F2F in Lund.