Minutes Globus integration task force 2011-07-14 15:00-16:30 CEST Minutes belonging to Agenda under https://www.egi.eu/indico/conferenceDisplay.py?confId=546 Presence: Michaela Barth MB Mario David MD Ioan Lucian Muntean IM Emmanouil Paisios (IGE, LRZ) EP John Gordon JG Steven Crouch SC Ilya Saverchenko IS Emir Imamagic EI Helmut Heller (IGE, LRZ) HH Tiziana Ferrari TF Peter Solagna PS Oscar Koeroo OK MB: Welcome everybody, thank you all for joining during summertime. We have a full agenda today. I hope nobody has something against that I start recording. I'll be on vacation next week, so if some time passes it is better to have some real records of the meeting. MB: 1) Last meetings minutes: https://www.egi.eu/indico/materialDisplay.py?materialId=minutes&confId=496 MB: The last meetings minutes have been sent out some weeks ago, I received some comments and corrections, so I assume they are fine by now. Any more comments, now? MB: I think we agree on last meetings minutes then. MB: 2) Going through the list of Actionpoints https://wiki.egi.eu/wiki/Globus_integration_task_force AP: SC to keep us updated on Grid-SAFE/accounting efforts within Globus. Especially: * Summary of the meeting between JR and Steven Booth on 110607 and the status of the LRZ testbed. * More exact time frames for moving the complete code to SourceForge and the release process afterwards. Howto and details on the release process. * Keeping us updated on Globus relevant UR fields analysis within the EMI ComputeAccounting working group. Update: See mail from SC on 110620. MB: The first AP was on Steve to keep us updated on Grid-SAFE/accounting efforts and we received a very nice email from him on the 20th of June. What I understood from this mail that all code should now be on SourceForge and would be regularly updated. And it was still unclear if the GridSAFE installation on the testbed is now completed? SC: As part of our obligation to EGI we wrote a document called IGE accounting mechanisms which we sent to them about two weeks ago, I think. This document gave a quick overview on GridSAFE, what it is, its architecture and its components and also about the plans for the testbed, source code availability, packaging, etc. I don't know how widely this has been circulated in EGI. Have you received that? MB: I haven't received it, no. JG: I haven't received it either. And as taskleader in two of the activities in accounting I think I should have received it. So, I'd appreciate if you could send me an offline copy, Steve. SC: No problem, what I could do is send it to the taskforce mailinglist, ok? I'll do that after the meeting. TF: Yes, Steve, I received this document through the policyteam. So because there were some things that were incomplete and I wanted to send feedback in a way that it could be improved and then circulated. Please, accept my apologies for having hold the document for a while. SC: Ok, shall I wait for your version to be sent to me? TF: Nono, just circulate the document, that's fine. It took just longer than I was expecting to be able to provide comments back to you, so apologies for that. ---> Update: SC to circulate the IGE accounting mechanisms document giving a quick overview on GridSAFE. SC: Ok, alright, thank you. So regarding the status of the LRZ testbed of GridSAFE: we are currently still working towards getting it operational. We've run into a few issues, because the developer Steve Booth has been on holiday and has also been working on other projects, so what we plan to do is to get that done in the next couple of weeks, in a few weeks. So basically at the moment, you can get access to the testbed, so I can sign you up to access to it, all I need is the DN, the distinguished name, so when it is ready you can test it. What we need to do next is to get it up running in operation. Once that's done all you need is a client, a RUPI client, a GridSAFE client and once I have signed you up for the testbed you can then submit testURs to GridSAFE. So that's going to be one of the next things we are doing. In terms of timeframes, it looks like it is going to be probably a few weeks, could be longer, because of holidays over August period, of course. But the end deadline for that, I suspect, could be September. JG: Steve, is the GridSAFE client different from the NGS RUS client? SC: Yes, it is at the moment, and one thing we want to do actually during August is test the NGS RUS client with GridSAFE and see to what extend it works and if it needs to be updated in some way. In theory being an RUS client it should work out of the box, but we can't be sure of that and have to test it. Then finally, there is also the release process as well, which is part of the action. We partially covered this in the accounting mechanisms doc, which I circulate. Once the code works on the testbed we can then make sure that the code on SourceForge mirrors what is on the testbed and then you can download it from SourceForge, you'll be able to build and install it and use it out of the box, that's the idea. And then if you wish to package it as part of an rpm, this can be done as part of IGE within in a future IGE cycle, in 2012. So that really does depend on whether we see that as a priority. So I guess that is an open question to the group. MB: I think that sounds good, but if we have nothing prepared so far we will need to get it into the UMD. Tiziana, maybe you know more what we have to do if we suddenly want to add something there that we haven't planned to before. TF: It depends on what Steven mentioned: the priorities. If you want that code delivered, then the best thing would be to have that as part of an IGE release, then automatically validated and going through staged-rollout according to the existing processes. So we first have to discuss together, if we want it as part of the release, and if we have consensus on that then we do that. SC: That's it exactly. JG: So there are two issues: Firstly there is collecting accounting records, collecting your data for Globus sites, so that probably is quite high priority, the second one would be to pass Usage Records on to the central repository will take longer. But in fact if you are already collecting the records, then eventually when that is all done you will publish the historical records and will have them alltogehter there in the end. So the other stand of working that is how a GridSAFE repository then passes this information on, publishes it to a higher level repository, and last week or the week before we sent out to a list of other accounting systems information about our new APEL central repository or at least of the test version and what you needed to code against, that was assuming using ActiveMQ messaging to publish stuff which is our plan for the rest of this year. When we will able to take on an onward RUPI ingress I don't know. Probably in the next year. SC: Right, so about putting the GLUE in some aspect to take the records out of our repository and upload them properly into the APEL in some fashion, like you say. JG: I thought GridSAFE did that, I thought it was designed to be hierarchicle. I'd have to doublecheck again, but we certainly had the intention in the UK for a GridSAFE instance to publish to the RUS at Manchester. SC: Right, ok. So that is something I can have a look into and confirm. JG: If you got that far then either we can provide a standards, you know a RUPI, interface to receive that or we can do like we do at Manchester. The Manchester one is just a prototype, sorry: the RUS part is in production, the passing on to APEL is just a prototype, it doesn't pass on to the production database, it just goes into a test database at the moment. What we could also do - if it is easily modular - then we could change that publishing bit once you've got the records and extracted them the way you should do. Then we could look at publishing by UR over ActiveMQ for example, just taking the records as the come out of the database. That is what our testsystem does. And other people at the moment. SC: So, if you had a RUPI endpoint sitting at your end then obviously you could do the integration with ActiveMQ and APEL there. JG: Yeah, if we had a proper RUPI interface it would be alongside ActiveMQ. But the other point I was thinking about is: that if we want to keep the records, at the moment we are keeping this archive in our peers NCSN UR schema, if we wanted to accept XML documents, we could also accept the XML documents over ActiveMQ. There are several different options we could work with and maybe the different people use the different options. And maybe in the longterm everyone uses standard OGF webservers. SC: Right, ok, that makes sense. So a final question is, if we would target, say for example early September for the preliminary work, would that be sufficient? Getting the testbed operational would be the main thing at that. JG: The preliminary work was what? For Globus sites to be able to publish to a GridSAFE or to have a testbed GridSAFE that people could try against? SC: Exactly, if we would target September for that, would that be sufficient, would that be a good time? JG: I have no opinion, but I will ask the UK Globus sites, that we have. SC: Ok, thanks. MB: So as I understand it, we have to ask if we really want to have it as part of an EGI sanctioned release? Should we already discuss and decide on this now or should we wait until early September for the first test? SC: I guess, that's entirely up to you. TF: Wait, I understand that this setup has first to be tested and not all necessary functionalities are available at the moment, is that correct? Because as soon as we have something that is part of the release I think this is added value for EGI for the software is ready for deployment and to pass verification and staged-rollout and if this is not the case and the software is more a prototype, then maybe.. JG: Nono, the software is claimed to be working, but we have not actually seen that yet. So I think having a seperate testbed that people can test against seems the best idea. The deadline for the next stage until after this probe. SC: Yes, that was the motivation for originally having installed it on the testbed for that reason. TF: So if end of September is the timeline for having the testbed set up and having the test running, then we probably should postpone a decision until we have the results of those test campaign, right. SC: Yes, that makes sense to me, and also in addition to that would be nice for our planning purposes for the next development cycle, so we'd know early enough by that point if we are going to include it in the third develpment cycle 2012 so we can package it and supply to EGI. TF: Steve, can you expose the timelines for the future release of IGE, what are more or less the windows when the software has to be tested, refined and the time needed to include that as a formal part of the IGE release? SC: The third development cycle still has to be decided on the dates, I think. Helmut can you comment on that? HH: The third development cycle will be 6 months after the second, the second is early 2012, say January, so that would be something like July of 2012, right Steve? SC: Sounds right to me, yeah. TF: And what you are talking about is the development cycle, correct? SC: This would effectively be packaging the component into a future IGE release, yes. TF: So I guess, if the testing is completed by the end of 2011, which is even less ambitious than what you said before that it should be completed by the end of September, then this would give IGE 6 months to do the packaging and have a release later on. Are also the release dates known, you mentioned a cycle from January 2012 until June? What does it mean? If the software passes the testing and is packaged at the beginning of 2012, that it is not released until July 2012? SC: That would originally be the case, yes, fitting in with the IGE release cycle, would that be too late then? TF: It really depends on the results, but if the software is mature enough, and I would guess that you probably have a good feeling of that, then this looks like we need to wait for like 12 months to have it deployed, right? SC: It depends how we do the development cycle, of course. So we have to decide on how this should be done. I am not sure if there is any lee way for modifying this cycle for the sake of GridSAFE. But there is a possibility we could squeeze it in the next IGEv2 release. But we'll have look into that and get back to you. TF: So there are no plans for intermediate releases earlier, I mean every three months or so. SC: No, that's correct. MB: So I noted that we'll have a decision then when the testbed is complete, whether we should have it as a part of an EGI/UMD release. I don't know Tiziana what we have to do technically for it to be part of the release, do we have to do something special, or is this then just part of the UMD and IGE release processes? TF: We don't need anything special. IGE Globus components are already planned for release in September, if I am not wrong. Having one more component that is validated and passes staged roll-out is no problem. All the processes are there, as soon as we think that this is ready for deployment and packaged by IGE, nothing changes from an EGI point of view. MB: Ok, great. MB: I think we can go on to the next AP. Any more comments to this AP? 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. * AP: TA to ask within NGI-DE for a second EA volunteer. Update: LRZ will be an EA for Globus. AP on MB to send a reminder to all NGIs to tell their sites to register all their Globus GT5 services in GOCDB, since this is a good time now with the upcoming SAM/Nagios release. MB: Next AP was the desired timelines for Globus integration, we had a small survey and this was also connected to the need for a second EA team. Last time we said that LRZ will be an early adopter for Globus. I've sent out an reminder to all NGIs to tell their Globus GT5 services to register in the GOCDB and Torsten wanted to ask within NGI DE for a second EA volunteer. Mario thinks that there is a need for a second EA volunteer. Torsten isn't here today. Helmut do you happen to know if Torsten sent out this request within Germany? HH: I would have to check my email, I can't say out of the top of my head right now. MB: Mario, would you like to elaborate on why you think a second EA would be needed? MD: Hi, at the moment it is sufficient to have one EA for Globus. We are always seeking for more than one for measures of redundancy, there will always be times when one person is not able to do it, and also in terms of different deployment scenarios that some sites have, when there are different settings for the same components. That is basically the main reason. But one EA is sufficient. As of now, for a couple of weeks the Globus components have been part of the process. They have been sitting still because we have been occupied with work from the EMI components. In a couple of weeks we'll have a first update to the first UMD release, but the tickets are opened and soon, after we've passed this stressful phase of UMD releases, we'll take care of the Globus components. So it will be rather soon, in one or two weeks, we are planning to make the verification of the components. After that the Early Adopters are expected to see notification for staged-rollout for those components. This will end up in August, which I don't know with the plans for holidays and all that. But in the end the target date for UMD 1.2 is in the middle of September, even with vacations now in August this should be feasible. Basically that's it. I would also like to ask the following: As I understood the account records are being done, I mean, it is still to be checked, if I understood correctly, to properly publish the accounting records in the APEL through ActiveMQ, even if it is released in the EGI, in the UMD, there are still at least if sites deploy they won't yet publish the accounting records, is this true? HH: true. And I wanted to say LRZ is ready for the flood of tickets from you and August is no problem, we have a couple of people here and we hope it is a quite time so that it is actually perfect to do that. MD: Ok, fine, great. The second thing which is in this AP is that there is some Nagios... TF: Mario, can I say something, can't we just have a second site in the UK? Because David Wallom was very keen on having Globus sites integrated in the UK. It makes sense, what do you think? JG: Yeah, I'm sure I can find a volunteer.So, we are talking about the IGE release in the UMD, we are talking about staged-rollout, right? So we have sites already and they are running pre-webservices GT4 I think at the moment, but I'm told that is not a difficult transition, so I'm sure I'll get at least one of them to volunteer. MD: Ok. Just send me a mail with information that is in the usual early adopters table and that's it. And from there on when the components go through staged-rollout they will be notified to do the staged-rollout of those components. [15:25:10] Ioan Lucian Muntean UTCN (as part of Romanian NGI) plans to deploy GT5 in summer 2011 JG: Of course, I remember most traditional staged-rollout sites are production sites. These sites have not been in production before, they are uncertified sites, that won't we an issue? MD: No, that's not an issue because presently there are no sites in production with Globus. We of course advise that the service should be in production, but sometimes they are test machines they are integrated gradually into the production, why not putting the site in production and... TF: Mario, is the site John is thinking about a regional one, which will never be integrated in EGI? JG: No, it is a regional one that can't be certified because it is not passing the SAM tests. But if it passes the new SAM tests it will be integrated into EGI, definitely. TF: Then Mario how can you do staged-rollout if the site is not a production one? To me it doesn't make a lot of sense without passing the monitoring. If you want to check the availability of the services how would you do that? The site has at least to be in the monitoring system. MD: That was going to be my question: Would this upcoming SAM/Nagios release that has been talked about in this Actionpoint, which release of Nagios is this? Is it the upcoming release 12 now? MB: It is already now part of SAM release 11, Emir is here so he can confirm, but the Globus probes should already be in SAM release 11. EI: The probes are part of the release. MD: But not yet in the critical set.. EI: Please, stop using this expression, it is confusing, but you are right they are not yet in the operations set. Because someone needs to request for them to get into the operations set. I don't know if we have a single Globus service monitored with any of the available probes. MD: Probably this is a thing that can start to be done, I mean the probe has to be tested, I mean the site has to have this thing functioning. So that you can evaluate when the probe can pass the critical tests. EI: The problem here is that in order for something to become an operational test, it has to pass procedure what is the number again, it has to be in OK state on most of the sites. And if we don't have any site which is tested with this thing, we won't be able to fulfill this requirement. MB: I have this later in the agenda, too BTW. We could just update this actionpoint and go on with the agenda. [15:35:35] Emir Imamagic https://wiki.egi.eu/wiki/PROC06 HH: I think we have some bootstraping problem here with Globus, since it hasn't been in there before. Probably some rules have to be applied to actually get it in. MB: That was just a question about the early adopters, Mario, we have an own point for staged-rollout later. Let's have the complete picture before we start to discuss. EI: I agree we can discuss it later. MB: We'll go on with the next actionpoint. For this actionpoint with the EA, we now just update that John will ask a site in the UK to volunteer. JG: Yes, I can do that, can I just ask for clarification,.. [audio problems for MB who didn't hear anybody] MB: .............................. 15:35:48] Emir Imamagic michaela [15:35:52] Oscar Koeroo Michaela [15:36:00] Mario David hey... Michaela ??? [15:36:16] Ioan Lucian Muntean I just wanted to add that [15:36:28] Ioan Lucian Muntean in Romanian we plan to have a GT operational site [15:36:29] Mario David can you hear us??? [15:36:49] Helmut Heller michaela - can you hear us? [15:37:02] peter solagna Michaela it seems that you can't hear any of us MB: I will try to restart. [some part of the discussion was lost, and from here on the minutes were not based on records, but on memory:] TF: Couldn't we have the sites certified, probably for that we will have a set of special operational tests. Once we have that available, once that is complete, at that time already the Globus release catches up. We could have this at NGI level. MB: Yes, we clearly have a chicken and egg problem here. We can't certify sites without a set of operational tests, and we can't make the tests operational without any testing certified sites confirming that the probes work. JG: Are we happy that the probes work? EI: They sites still have not taken them over, but we need to wait for feedback. IS: I tested them locally and they were fine, but of course to test them in a larger environment wouldn't hurt. EI: All of these probes have been used before, noone complained then. I don't think deploying them on a larger scale will lead to different results. JG: Can Mario see the results of the probes even if they are not operational yet? EI: All the results go to the central database, in theory Mario as a dteam member can go and see the results at the national instances as well. MD: The sites install and deploy the software when we notify them, the thing that is missing really is the monitoring, but we can let the staged rollout go on for the moment to see how things are going. MB: Agreed. MB: Going on now with the next open action point about Operating Systems:: * AP: (on request of COO after a discussion at the TCB): Which type of OS for GLOBUS resources will be used by the NGIs that plan to integrate their GLOBUS resources into EGI? This information is important for the integration of GLOBUS into the UMD. Update: "if we do not have sites that volunteer for EA activity in StagedRollout then there is no point in including that particular platform (e.g. CentOS on x86_64) in the UMD!" Update: AP on MB to update Michel Drescher on the OS discusion we had at the meeting 110606. Done 110609. Update: Currently just SL5 supported by UMD. Requests on further plattforms should go over OMB to TCB. Every supported plattform increases workload for verification and staged-rollout. OK: In IGE we see a diverse set of plattforms, but that doesn't flow into EGI, which is also fine. MB: As I tried to say: We are just a small set of NGIs and every additional plattform would need its own Early Adopters with additional verification and staged-rollout. So at the moment it sounds reasonable to just concentrate on one OS, we are already familiar with within EGI, and when everything works in staged-rollout to take up this actionpoint again and request this officially through the OMB and the TCB. Until then I suggest we freeze in this actionpoint. Can we agree on this? --> freezing this actionpoint. MB: * AP: MB to circulate link of MS409 as soon as it comes out to answer questions about the new staged-rollout process. Update: https://documents.egi.eu/secure/ShowDocument?docid=478 as in mail of 110613 Mario would you like to add something? Has anybody questions to Mario about this right now? MD: This document is almost final and currently undergoing AMB review, so the procedures for the staged roll-out can be considered valid. You find them either here in this document or on the wiki. Any questions now or later via email are welcome. MB: * AP on Nagios Probe support responsibilities: HH to answer Emirs email on IGE officially taking over probe support in the future. The details should be fixed within a month. Has this mail been sent now? EI: I haven't received anything. HH: I think I have sent it, Steve, didn't I tell you that I sent the email? Maybe it got lost in transaction. TF: What is this AP about? MB: It is about IGE officially taking over SAM/Nagios probe support. TF: But that is surely already in the minutes and we all heard it. EI: It is just fixing the technical details and the names. HH: I'll send you the mail again. --> AP on HH to send this mail again. MB: * AP: MB to ask OK for a nice summary of status of Globus integration into Argus at the next meeting. has been done, we'll have that later in the agenda. MB: * AP: MB to create a doodlepoll for the next meeting in the second week of July. If another meeting is necessary before the TF in Lyon will be decided during this meeting. Update: http://www.doodle.com/dk532e78uerp84nh Ok, the decision for another meeting is its own point later in the agenda, as well. 3) Discussion of Progress: a) GOCDB MB: I sent out the request to the NGI managers to enter their Globus sites with GT5 services into GOCDB, can I ask the people here presenting an NGI if there were any problems? How many sites with Globus GT5 services are now in GOCDB, just LRZ and Dortmund? TF: Just Dortmund has real Globus GT5 services in GOCDB. EI: There was also a Finish site which registered a Globus GridFTP to get it to work instead of an ARC GridFTP, which led to an error, but that was just a work around. [15:58:20] Emir Imamagic se2.fgi.csc.fi [15:58:29] Emir Imamagic https://ping.fgi.csc.fi/nagios/cgi-bin/status.cgi?hostgroup=node-globus-GRIDFTP&style=overview TF: John, could you contact Dortmund and try to have some accounting tests with them against your new test system? JG: Yes, I can do. -->*new* AP on JG to ask Dortmund if they are able to make some accounting tests against the new APEL test system. MB: b) Monitoring https://rt.egi.eu/rt/Ticket/Display.html?id=390 So the ticket has been closed and starting with Update-11 Globus probes are integrated into SAM: https://tomtools.cern.ch/confluence/display/SAMDOC/Update-11 One thing I wanted to ask Emir is that in the ticket in the end, you start to discuss the possibility to use the (gLite) BDII with Globus services, answering a question of Foued. Could you elaborate on that a bit, that sounded interesting? EI: Oh, that is an ugly hack actually. A real information system should be used instead. It wouldn't go through staged-rollout like this. And it has problems with VO and closeSE. HH: This would fulfill about the same task as the future Globus IIS? IS: The Globus 5 information system is not available at the moment, from my point of view. So it is a question of what is the timeframe, what is the most reasonable thing, do we need a temporary workaround? MB: Almost all the procedures, like the site certification procedures for example require to have a kind of BDII, so if we could have an easy way to broadcast site information... TF: The IIS is not yet in production, as I understand? MB: Ilya, It is still in alpha state, right? IS: It won't be ready for production until the end of the year. MB: This is why I was interested in this work around, Emir. EI: This is basically a hack. Someone from either SA2 or the actual site admins would have to check throughly if this would fit into the rest of the BDII information generated. TF: Emir, you said it won't pass staged roll-out, why not? EI: I could provide my notes, it is not a documentation and someone can look at this. As I said Globus doesn't have this concept of closeSE. If you have any site admin or Mario David would possibly be a good candidate to look at this. IS: Functionalitywise the IIS is quite ambitious. If there could be a document available describing what is exactly expected, it could be that the current version of IIS, even though still in alpha stage, is already cabable of generating such information. What is needed? JG: Basically, it is the GLUE schema. Depending on the version GLUE1 or GLUE2. MB: For ARC you have such a glite site-BDII running where an internal script transforms the GIIS information into BDII speech, so to say. EI: You can literally reuse all the things, the only problem is the CloseSE. But, you can point it to something that is a classic SE. The output is literally the same. IS: Are you providing dynamic information also in the case of gLite? EI: Since we use the gLite information provider, you get all the information. If you find the gLite UI you can execute something like this. EI: This service is speaking GLUE we just utilize the information providers. JG: You say the Globus sites should run a gLite BDII since you can't get any of the other stuff. Can't you just pull some of the information providers? IS: What we need is a well defined "must" subset of the information that should be provided. IGE could then take a look at it and try to fulfill that subset of the GLUE schema. EI: Essentially you just need the CE stuff. EI: How do you build a scheduling information on top of the IIS? IS: You would have all kind of static information by default, and dynamic information you will be able to distribute, like via Nagios and Ganglia. EI: Sounds like it is providing the same functionality as a WMS. JG: I suggest that you don't try to reinvent the wheel. gLite has all this functionality how to interact with the scheduler and so on. EI: Yes, this is why we used that part from gLite. You can put anything above it, and you don't have any conflicts there. It was one year ago when we decided to go with this solution. MB: I think we would need a volunteer to test this and evaluate if this could be a working solution. A volunteer who really tests this. And Emir would have to provide some more technically detailed info. IS: I could possibly be such a volunteer. I would of course have to coordinate with Helmut and other people from IGE first. Emir: Ilya, if you write an email to the list or to me personally, I'll give you the technical details. IS: I'll most probably will write an email to the list. MB: Good, then we continue this discussion on the mailing list. [16:08:00] Emir Imamagic mon.cro-ngi.hr [16:08:34] Emir Imamagic lcg-infosites --bdii mon.cro-ngi.hr --vo crongi ce MB: We drifted a little bit away in the discussion. Any more problems with monitoring? Do we consider the probes to work fine? They are not really tested yet, yes? EI: We consider them to work fine and expect no problems. MB: c) Staged roll-out Tickets in the sw-rel for globus (verification phase) https://rt.egi.eu/rt/Ticket/Display.html?id=2436 https://rt.egi.eu/rt/Ticket/Display.html?id=2441 https://rt.egi.eu/rt/Ticket/Display.html?id=2446 https://rt.egi.eu/rt/Ticket/Display.html?id=2451 https://rt.egi.eu/rt/Ticket/Display.html?id=2456 https://rt.egi.eu/rt/Ticket/Display.html?id=2459 EAs to get in contact with Mario David davidATlip.pt We discussed this now already in the Actionpoints. Here you find a link with the current tickets in the verification phase, where Mario said they would be taken care of in one and a half week and we can then possibly start the staged-rollout process with the Early Adopters early August, and see how far we come then. MB: d) Accounting There we have seen a last mail answer from John Gordon end of June. Has anything happened since then, would you like to add something John? JG: Which mail, I don't remeber exactly, but definitely, I would like to add something. This has already been brought up earlier: we have a new test system at RAL, which can be used to test against. MB: What about the workshops you mentioned previously, is there already something organized now. You said you would contact the relevant people and invite them. JG: Yes, we will have dedicated workshops at the technical forums. I already sent out emails to the developers asking for their input. Steve, did you get one such mail? SC: Yes. MB: So, Steve should then contact you and give you a list of his wishes what should be adressed in these workshops, right? JG: Correct. MB: I will forward this information also tomorrow in the UNICORE integration task force. JG: And Michaela, if we could have the accounting point a little bit earlier tomorrow in the agenda, please. MB: Of course. MB: 4) Authorization a) Update on Argus Oscar has volunteered to give us an update on the status of Globus security and Argus integration. OK: Our current plans for the moment are to create a preview release for internal use, for technological review only, to interconnect all the GT5 services in the Argus system. The following features concern the Globus gatekeeper, gridftpd and gsi-opensshd. OK: Features to ramp up to Argus integration, planned to be released as part of UMD 1.2: - Non-VOMS poolaccount support (legacy feature) - VOMS-based authorization and (pool)account mapping OK: Based on this, we'll give a roadmap on how a site can upgrade. We now have the pool account and VOMS mechanisms, but afterwards it will soon also include direct Argus support. OK: After UMD1.2 there needs to be a small rework into some components, afterwards we'll be adopting direct Argus support, and adding more support plattforms. From there on we can make small adjustments until the end of the year. OK: Feature planned before the end of the year: - Integrate the Argus call-out as a supported plug-in OK: On the todo list: - Minor development in the already existing Argus plug-in - Ensure that the Argus protocol libraries are suitable for integration on the platforms IGE wants to be able to deploy on. Some issues need to be resolved for SLC6/CentOS6 and probably Debian6 too. MB: So I understand that with UMD1.2, we'll already have VOMS support with Globus, right? OK: correct. MB: Any more questions to Oscar? MB: 5) Timelines and next procedures to take into account a) Changes needed from Globus point of view: to OLA? https://documents.egi.eu/secure/ShowDocument?docid=31 to site certification procedure? https://wiki.egi.eu/wiki/PROC09 b) Nagios tests going operations https://wiki.egi.eu/wiki/PROC06 c) Switching on availability calculation https://wiki.egi.eu/wiki/PROC08_Modification_of_the_set_of_AVAILABILITY_tests I would like to ask you all to have a look at the OLA and the site certification procedure, to see if you see any problems when applying them from a Globus point of view. Because this will be one of the next things we'll have to go through will all Globus sites. The OLA as well as the site certification procedure were tried to be written as middleware independent as possible. Then we already touched the Nagios tests going operations procedure and the swichting on availablity calculation procedure. Here just that you are aware of them and know were to find them. I don't expect any problems when applying those two. MB: 6) Input needed for MS414 Integrating Resources into the EGI Production Infrastructure https://documents.egi.eu/secure/ShowDocument?docid=650 Similar to last year with MS407 we'll have also this year a milestone with updated info on how to integrate resources into the EGI Production Infrastructure. I'll ask some of you for input or already have, and some of the material that came up during discussions in our task force will also be taken as input for this milestone and me or Gert will ask you to review this milestone. MB: 7) Taskforce leadership handover Gert Svensson gert@pdc.kth.se will take over from 1st of August. Gert will take over during the next 10 months while I am on maternal leave. So this will be my last meeting with you. He is not here today because he is on vacation, but we had a transition period of several months and he has listened in to many of our task force meetings and I am sure you will all help him in his new position. TF: Thanks for your efforts during Project Year 1 and good luck with your new responsiblities! All: Yes, good luck. MB: Thanks :-) MB: 8) Next meetings a) fixed at the EGI TF, Lyon, week 38 b) one more meeting during August? With all the discussion we had now here during the day, maybe another meeting in August would be a good idea anyway? JG: There won't be happening much during August. EI: Besides Staged-rollout. There we would eventually need another meeting. MB: So you suggest a dedicated meeting just for staged-rollout? Mario, what do you say? MB: Mario is greyed out now. I think we leave it up to him, whether he thinks a dedicated meeting focusing around staged-rollout is necessary. MB: 9) AOB MB: Ok, then thank you for joining and sorry that we ran out of time. Good bye! ----------------------------------------------------------------------- Hi Michaela, I tried in the meeting several times to have the word on the second action point, but i was not successful with that. Please put in the notes our status: At UTCN we have a production site (g-Lite based), registered with our NGI, on which we plan to deploy in summer 2011 GT5 components, and thus, act as a stage rollout site. Thanks & cheers, Ionel. ------------------------------------------------------------------------ Updated Actionpoints after the meeting: * AP: SC to keep us updated on Grid-SAFE/accounting efforts within Globus. Especially: o Summary of the meeting between JR and Steven Booth on 110607 and the status of the LRZ testbed. o More exact time frames for moving the complete code to SourceForge and the release process afterwards. Howto and details on the release process. o Keeping us updated on Globus relevant UR fields analysis within the EMI ComputeAccounting working group. Update: See mail from SC on 110620. Update 110714: SC to circulate the IGE accounting mechanisms document giving a quick overview on GridSAFE including archictecture, components, planned release process,... People can send their DN to SC to access and test the testbed. Current Deadline to get GridSAFE testbed fully operational: September. After that code on SourceForge will be updated to mirror this working version. In August: Plan for a comparison between GridSAFE client and NGS RUS client (any RUS client _should_ work out of the box). When testbed works: SC will check UK plans to publish from GridSAFE to the RUS in Manchester (RUS part in production, passing on to APEL in prototype stage). JG will ask the UK Globus sites. Postponing a decision on whether GridSAFE should be part of an IGE release taken into account by UMD until after the testbed is fully operational. IGE development cycles: second January 2012, third July 2012. Actual deployment might take a long time. Investigate possiblity to sqeaze it into IGEv2 instead. No plans for intermediate IGE releases. * 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. Update: LRZ will be an EA for Globus. AP on MB to send a reminder to all NGIs to tell their sites to register all their Globus GT5 services in GOCDB, since this is a good time now with the upcoming SAM/Nagios release. Update: JG to ask for EA volunteer within the UK, UTCN in Romania volunteering as well (g-Lite based production site which plans to deploy Globus GT5 components in summer 2011). Contacts and Informations should be sent to Mario David. Identifying possible problems due to a still missing operational set of SAM/Nagios Globus probes during staged-rollout as we go. * AP: (on request of COO after a discussion at the TCB): Which type of OS for GLOBUS resources will be used by the NGIs that plan to integrate their GLOBUS resources into EGI? This information is important for the integration of GLOBUS into the UMD. Update: "if we do not have sites that volunteer for EA activity in StagedRollout then there is no point in including that particular platform (e.g. CentOS on x86_64) in the UMD!" Update: AP on MB to update Michel Drescher on the OS discussion we had at the meeting 110606. Done 110609. Update: Currently just SL5 supported by UMD. Requests on further plattforms should go over OMB to TCB. Every supported plattform increases workload for verification and staged-rollout. Freeze: Taking up this actionpoint again when everything works in staged-rollout and discuss an official request through the OMB and the TCB. --> on hold. * AP: MB to circulate link of MS409 as soon as it comes out to answer questions about the new staged-rollout process. Update: https://documents.egi.eu/secure/ShowDocument?docid=478 as in mail of 110613 Update: MS409 almost final and currently undergoing AMB review, so the procedures in it for the staged rollout can be considered valid. You find them either here in this document or on the wiki. Any questions now or later via email to Mario David are welcome. --> closed. * AP on Nagios Probe support responsibilities: HH to answer Emirs email on IGE officially taking over probe support in the future. The details should be fixed within a month. Update: AP on HH to resend the mail with the technical details and the names. * AP: MB to ask OK for a nice summary of status of Globus integration into Argus at the next meeting. Update: A status summary has been given --> closed. * AP: MB to create a doodlepoll for the next meeting in the second week of July. If another meeting is necessary before the TF in Lyon will be decided during this meeting. Update: http://www.doodle.com/dk532e78uerp84nh . Update: If we need an other meeting during August then a dedicated one about staged-rollout: AP on MD to decide whether we need a dedicated meeting on staged-rollout during August. *new* AP on JG to ask Dortmund if they are able to make some accounting tests against the new APEL test system. *new* AP on EI: to share his notes and technical details on the gLite based BDII hack for Globus. IS to eventually test this and probably compare with IIS (to derive a basic set of what is needed to be compatible?).