15:31:03 <lalatenduM> #startmeeting StorageSIG 15:31:03 <centbot> Meeting started Fri Jan 9 15:31:03 2015 UTC. The chair is lalatenduM. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:31:03 <centbot> Useful Commands: #action #agreed #help #info #idea #link #topic. 15:31:12 * hchiramm in 15:31:23 * kbsingh is here too 15:31:31 <lalatenduM> who we have here for the Storage sig meeting? 15:31:43 <bvanassche> Bart Van Assche 15:31:48 <scuttlemonkey> <- 15:31:54 <hchiramm> Humble Chirammmal :) 15:31:57 <lalatenduM> #chair billings scuttlemonkey bvanassche kbsingh hchiramm 15:31:57 <centbot> Current chairs: billings bvanassche hchiramm kbsingh lalatenduM scuttlemonkey 15:32:17 <lalatenduM> anyone I missed? 15:32:25 <lalatenduM> alphacc, ^^ :) 15:32:44 <lalatenduM> Happy new year folks :) 15:33:04 <lalatenduM> #topic Agenda 15:33:12 <lalatenduM> #info Topic: Status Updates 15:33:12 <lalatenduM> #info Subtopic: GlusterFS 15:33:12 <lalatenduM> #info Subtopic: OpenAFS 15:33:12 <lalatenduM> #info Subtopic: Ceph 15:33:12 <lalatenduM> #info Subtopic: SCST 15:33:13 <lalatenduM> #info Open Floor 15:33:41 <lalatenduM> #topic Action items from last meetings 15:34:03 <lalatenduM> I missed to put this topic in the agenda :( 15:34:16 <lalatenduM> anyways lets go through the AIs 15:34:33 <lalatenduM> billings to build OpenAFS using koji 15:34:47 <billings> I just got my koji credentials this morning 15:34:50 <lalatenduM> billings, lets carry it forward to next meeting 15:34:51 <lalatenduM> yeah 15:35:02 <billings> thanks 15:35:10 <lalatenduM> billings deason alphacc kbsingh to to brainstorm on some infrastructure/automation to automatically build a new srpm every time a new kernel comes out 15:35:22 <billings> could you make that an action? 15:35:29 <lalatenduM> #action billings to build OpenAFS using koji 15:35:34 <lalatenduM> done 15:35:34 <billings> thanks 15:35:49 <kbsingh> from the reimzul side - where the distro components are built - we already have trigger support 15:36:00 <billings> excellent 15:36:05 <lalatenduM> awesome 15:36:13 <kbsingh> so how i see that working in this requirement is : srpm is imported, kernel does its thing, built and pushed into the bstore ( that is the justbuilt ) repos 15:36:14 <billings> there are some issues I have with the current storage7 targets 15:36:25 <kbsingh> that part is the easy bit 15:36:54 <kbsingh> the tricky bit is - how then do we make this bstore repo available to koji, immediately, then have the kmod's built against that repo 15:37:05 <kbsingh> once the kmod's are done - how do we test them to get some level of confidence 15:37:27 <kbsingh> finally, if all is good, we can release things in sync ( it will actually be kmod's first, then kernel ) 15:37:30 <billings> is the bstore repo a regularly named repo? 15:37:40 <billings> could we have a custom target that always includes it? 15:37:58 <kbsingh> note: doing this entire proces, the kernel isnt released... so it cant take too long - also, when something fails, do we hold kernel till its fixed, or do we time out at say 4 hrs ) 15:38:09 <kbsingh> bstore repos are internal to reimzul 15:38:18 <billings> an 15:38:20 <billings> ah 15:39:10 <lalatenduM> kbsingh, reimzul is CentOS core's build system, right? 15:39:30 <kbsingh> billings: http://buildlogs.centos.org/c7-updates/dracut/20141103134735/ is a bstore repo 15:39:39 <kbsingh> lalatenduM: yeah 15:39:42 <lalatenduM> #info reimzul is CentOS core's build system name 15:39:46 <billings> ah, so they're not staticly named 15:39:54 <kbsingh> so its not connected to koji itself, but it can run koji commands from the command line. 15:40:33 <kbsingh> billings: the datestamp bit is the UID for the build, it allows us to rebuild the same name-EVR repeatedly, the reimzul repo collector will only use the latest content ( by sorting on the UID ) 15:41:05 <billings> well, the point is that we couldn't use it in koji since koji would never know about it. 15:41:45 <kbsingh> right, so we need to push that somewhere and trigger a koji repo update thing 15:41:45 <billings> unless reimzul had the ability to *create* a target that included that repo 15:42:13 <kbsingh> from my side of the fence, i think we can find a way technically to make this work 15:42:30 <kbsingh> however, the thing that is a bigger challenge is going to be the policy side of things 15:42:42 <kbsingh> what do we do with the distro kernel if say, the openafs kmod failed 15:42:43 * billings nods 15:43:11 <billings> I have no problem having the policy be "if it builds, use it, if it doesn't, report and move on" 15:43:28 <lalatenduM> kbsingh, from storage sig point of view we should not release the distro kernel till openafs kmod builds fine 15:43:55 <kbsingh> so, is it just the openafs kmod that needs this ? iirc, the scst guys are ok to not need a kmod initially right ? 15:44:16 <kbsingh> lalatenduM: there is a way to make that work as well 15:44:20 <kbsingh> its a big of a hack 15:44:25 <lalatenduM> kbsingh, yeah or we take discussion at that moment 15:44:34 <kbsingh> but the workaround there would be to include the kernel in the openafs repo itself 15:44:39 <kbsingh> hve it linked from the distro tree's 15:45:05 <kbsingh> and then have the centos-release-openafs implement a yum-priorities to ensure only the kernel from openafs repo is used 15:45:21 <lalatenduM> kbsingh, ok 15:45:26 <kbsingh> that way, even if a newer kernel makes it to the updates/ repo, it wont be used on the machine of people who have the centos-release-openafs installed 15:45:34 <kbsingh> but hey, this is a proper hack and a half 15:46:02 <lalatenduM> kbsingh, lokks fine at this moment :) 15:46:24 <billings> yeah 15:46:33 <lalatenduM> I think that something people will signup for when they use storage sig 15:47:12 <billings> I could also make it an openafs-specific priority 15:47:12 <lalatenduM> so who owns the action item to try this out? 15:47:20 <kbsingh> note that if we do the yum priorities thing, then we need all the kernels in the openafs repo 15:47:21 <lalatenduM> billings, yes 15:47:38 <lalatenduM> kbsingh, yes 15:48:09 <kbsingh> right, and then someone should try this out manually and make sure its going to do what we think its going to do ( i am fairly confident, but its still something worth testing ) 15:48:43 <lalatenduM> billings, should I put the AI on you? 15:48:50 <billings> sre 15:48:51 <billings> sure 15:49:29 <billings> I'm concerned that the existing storage7 target I tested against seems to be missing any updates since c7 was released 15:49:45 <billings> so we're certainly going to have to fix that 15:49:54 <lalatenduM> #action billings to try out reimzul's way of kmod building and yum priorities stuff for openafs repo 15:49:58 <hughesjr> have we solved the openafs license issue (speaking of openafs) 15:50:12 <lalatenduM> kbsingh, ^^ 15:50:19 <billings> there's a licensing issue? 15:50:26 <alphacc> billings: Let me have a look. It may be due to priorities. 15:51:23 <lalatenduM> billings, honestly I dont know 15:52:23 <billings> I think there was some concern with CentOS packaging OpenAFS, but I think its ok as long as it's labeled with the IBM license. 15:52:32 <lalatenduM> hughesjr, may be this something you and kbsingh can discuss offline and update us 15:53:00 <lalatenduM> billings, lets take that offline 15:53:06 <billings> ok 15:53:08 <lalatenduM> moving ahead 15:53:15 <lalatenduM> lalatenduM will help deason to get standard policy or guidance on packaging selinux rules 15:53:18 <kbsingh> ok, we can work through the licensing stuff - i think it was mostly around it being GPL compatible or not and if it tainted kernel 15:53:22 <lalatenduM> deason, around? 15:53:42 <kbsingh> #action alphacc to validate the storage7 koji targets and setup new tags as needed for openafs 15:53:50 <hughesjr> billings: yes ... it uses the IBM Public License .. that is not GPL Version 2 or 3 compliant .. makes kernel modules problematic 15:53:54 <lalatenduM> #action kbsingh to see if openafs is GPL compatible or not 15:54:03 <billings> hughesjr: indeed. 15:54:27 <alphacc> billings: ok, it's fixed... 15:54:48 <lalatenduM> #action lalatenduM will help deason to get standard policy or guidance on packaging selinux rules 15:55:02 <lalatenduM> It is not done , hence moved to next week 15:55:12 <lalatenduM> lalatenduM will help scuttlemonkey on 101 for building Ceph on cbs 15:55:21 <lalatenduM> scuttlemonkey, we have not done this 15:55:32 <scuttlemonkey> yeah 15:55:32 <lalatenduM> may be because of holidays :) 15:55:36 <scuttlemonkey> nod 15:55:43 <lalatenduM> lets carry it forward 15:55:47 <lalatenduM> #action lalatenduM will help scuttlemonkey on 101 for building Ceph on cbs 15:55:55 <alphacc> billings: you did launch it a bit fast repo were regenerating ;) 15:56:00 <billings> haha 15:56:04 <lalatenduM> :) 15:56:08 <lalatenduM> kbsingh will start a discussion on "the build system should build the kmod packages from a spec file rather than having the makefile specificall build the rpm" as openafs and scst will need the functionality 15:56:10 <scuttlemonkey> lalatenduM: lets shoot for next wed if that works? 15:56:12 <billings> All I did was hit ^p enter, so it was easy. :) 15:56:24 <alphacc> effective multitasking 15:56:24 <lalatenduM> scuttlemonkey, fine 15:56:40 <kbsingh> ack 15:56:59 <lalatenduM> kbsingh, billings I think it is covered in the previous action item , right? 15:57:14 <billings> yes 15:57:19 <alphacc> billings: ok please ^p 15:57:34 <lalatenduM> vlnb to sync with lalatenduM to get access of buildsys 15:57:38 <lalatenduM> this is done 15:57:42 <lalatenduM> thanks to alphacc 15:58:02 <lalatenduM> hchiramm and lalatenduM to think more how tos/wikis for gluster in storage SIG. 15:58:15 <lalatenduM> hchiramm, you want to add something here? 15:58:34 <lalatenduM> hchiramm, has got access to wiki 15:58:40 <hchiramm> Still need to finalize on articles 15:58:48 <lalatenduM> ok 15:58:53 <hchiramm> however I have the wiki setup and a title ready to start 15:59:06 <hchiramm> will add contents to that and will come up with other articles 15:59:09 <bvanassche> Small correction: Vlad asked me to request build system access. 15:59:16 <bvanassche> I received access keys a few hours ago. 15:59:19 <lalatenduM> hchiramm, should I put an ai on you? 15:59:32 <hchiramm> please feel free :) 15:59:49 <lalatenduM> #info Vlad asked bvanassche to request build system access. 16:00:03 <kbsingh> bvanassche: have you been able to test the info ? are you setup to start throwing srpms at the targets / 16:00:11 <lalatenduM> #info bvanassche has got the build system access 16:00:29 <bvanassche> koji login works, but I have not yet had the time to try sending an SRPM to the build server. 16:01:03 <kbsingh> ok, there are some wiki pages with details on howto do that ( if you havent used koji in the past ) 16:01:19 <hchiramm> lalatenduM, please put an AI on GlusterFS articles 16:01:20 <lalatenduM> bvanassche, feel free ping me for this 16:01:23 <lalatenduM> #action hchiramm needs to finalize wiki articles for GlusterFS 16:01:27 <hchiramm> :) 16:01:27 <lalatenduM> hchiramm, done 16:01:28 <kbsingh> bvanassche: my understanding is that you are going to only build userland stuff and nothing for the kernel specifically. 16:01:35 <kbsingh> bvanassche: eventually we might need to build a kmod as well 16:01:47 <kbsingh> bvanassche: and all the code is GPLd 16:01:51 <kbsingh> bvanassche: is that an accurate summary ? 16:02:00 <bvanassche> kbsingh: currently SCST consists of two RPMS - one RPM with kernel modules and one RPM with user space software. 16:02:12 <bvanassche> The GPL license applies to the source code of both packages. 16:02:33 <alphacc> bvanassche: the kmod is a real kmod doesn't need a rebuild with each kernel ? (if kabi is not modified) 16:02:35 <bvanassche> The RPM with kernel modules will have to be rebuilt after each kernel update. 16:02:53 <kbsingh> bvanassche: ok, but are we going to try and build the kmod's now ? i seem to recall from the last meeting that it wasent needed perse, it added a few features only used in a minority of cases 16:02:56 <bvanassche> SCST uses several kernel symbols that are outside the kABI. 16:03:13 <alphacc> bvanassche: ok thanks for confirming 16:04:17 <lalatenduM> #chair alphacc 16:04:17 <centbot> Current chairs: alphacc billings bvanassche hchiramm kbsingh lalatenduM scuttlemonkey 16:04:28 <bvanassche> kbsingh: The kernel modules contain the core SCST functionality and hence are essential. 16:04:48 <bvanassche> What is not essential are the core kernel patches (scst_exec_req_fifo and put_page_callback). 16:05:28 <lalatenduM> hmm, we need to build kmods 16:06:17 <kbsingh> bvanassche: ok 16:06:25 <lalatenduM> bvanassche, as I remember we dont need rebuild the kernel though as of now 16:06:51 <kbsingh> so if billings is able to confirm that the priority thing is going to work - we might need to do something similar for the scst repos as well 16:07:00 <lalatenduM> #info SCST would need to build kmods 16:07:01 <kbsingh> the trigger to build kmod's is easy 16:07:25 <lalatenduM> ok, moving ahead 16:07:38 <lalatenduM> #topic GlusterFS status 16:07:55 <lalatenduM> We are waiting for pkg signing infra 16:07:57 <bvanassche> lalatenduM: indeed - we are not asking to rebuild the CentOS kernel. During previous meeting I promised that I would rework the scst_exec_req_fifo patch and that I would transform that patch into core SCST code. I made good progress with that work and will send it to Vlad for review soon. 16:08:25 <lalatenduM> bvanassche, nice! 16:08:59 <lalatenduM> bvanassche, thanks for the efforts :) 16:09:24 <lalatenduM> kbsingh, alphacc , do you guys have any update on pkg signing infrastructure? 16:09:46 <kbsingh> lalatenduM: not yet, it came up in the cbs meeting on Monday - and its been worked out. i believ we should have something in the next week or two 16:10:27 <lalatenduM> kbsingh, ok. Should I publish the unsigned repos to mailing lists ? 16:10:51 <alphacc> kbsingh, lalatenduM : I am testing sigul. More news next week. 16:11:09 <lalatenduM> alphacc, cool 16:11:33 <kbsingh> lalatenduM: we should do what we did for openstack, lets get the repos on buildlogs.centos.org- make sure thr eis some docs on the wiki - and then we can announce it to the devel list for further testing etc 16:11:35 <alphacc> (the yum plugin we use at work for kernel/openafs matching http://linuxsoft.cern.ch/cern/slc6X/x86_64/SRPMS/yum-kernel-module-1-5.slc6.cern.src.rpm) 16:12:17 <lalatenduM> kbsingh, ok lets do that than as we have initial docs ready 16:12:40 <lalatenduM> #info Pkg signing infra might be available in next two weeks 16:13:02 <lalatenduM> hchiramm, you have anything for Gluster 16:13:21 <hchiramm> no for now 16:13:45 <lalatenduM> #action lalatenduM to publish the unsigned repos to mailing lists. 16:13:56 <lalatenduM> ok moving to next item 16:14:05 <lalatenduM> #topic openafs status 16:14:12 <lalatenduM> billings, any thing to add here 16:14:22 <billings> one moment 16:14:34 <billings> I think that deason had some patches he was working on 16:14:53 <billings> but I haven't heard any more about them, so I'm moving forward. 16:15:05 <lalatenduM> billings, cool 16:15:16 <lalatenduM> #topic Ceph status 16:15:24 <scuttlemonkey> lalatenduM: not much to say on this one 16:15:39 <lalatenduM> scuttlemonkey, yeah lets work on the action item 16:15:43 <scuttlemonkey> yeah 16:15:54 <lalatenduM> #topic SCST status 16:16:11 <lalatenduM> bvanassche, anything else you want to add? 16:16:34 <bvanassche> I think all I wanted to share has already been mentioned ... 16:16:45 <lalatenduM> bvanassche, yup. lets move ahead 16:16:53 <lalatenduM> #topic openfloor 16:17:01 <bvanassche> BTW, would it help if I would attend the CentOS meeting during FOSDEM ? 16:17:13 <bvanassche> I'm living in Belgium so that makes it easy for me to attend FOSDEM. 16:17:40 <lalatenduM> bvanassche, me and kbsingh , alphacc are coming to Fosdem and to the Dojo too 16:18:02 <lalatenduM> anyone else coming to Fosdem? 16:18:20 <lalatenduM> bvanassche, I think it will help 16:18:24 <lalatenduM> if you come 16:18:43 <bvanassche> OK, I will do my best to attend the CentOS meeting. 16:18:51 <lalatenduM> scuttlemonkey, are you coming for Fosdem? 16:18:55 <lalatenduM> billings, ^^ 16:19:02 <scuttlemonkey> lalatenduM: no, I'll be out for that one unfortunately 16:19:02 <lalatenduM> hchiramm, ^^ 16:19:09 <lalatenduM> scuttlemonkey, ok 16:19:33 <lalatenduM> ok. moving ahead 16:19:42 <kbsingh> bvanassche: come to the Dojo on friday 16:19:51 <bvanassche> OK 16:19:52 <billings> I am not going to fosdem. 16:19:53 <kbsingh> also, i think Loic is doing something on Ceph at the dojo 16:20:06 <billings> Brussels is a bit of a hike from here. 16:21:10 <scuttlemonkey> yeah, Loic is planning a demo/walkthrough type presentation 16:21:12 <lalatenduM> billings, where do you stay? 16:21:28 <lalatenduM> scuttlemonkey, thats in Dojo or Fosdem? 16:21:40 <lalatenduM> ah, dojo 16:21:42 <scuttlemonkey> it's part of the dojo 16:21:43 <scuttlemonkey> yeah 16:21:44 <lalatenduM> got it 16:21:50 <lalatenduM> ok moving on 16:21:54 <lalatenduM> Planning to have the next sig meeting after 2 weeks i.e. on 23rd Jan. Does it work for all? 16:22:21 <scuttlemonkey> yep 16:22:32 <kbsingh> yeah 16:22:42 <kbsingh> we can all syncup on the maillist in the mean time 16:22:58 <billings> fine. 16:23:03 <lalatenduM> #info The next sig meeting is after 2 weeks i.e. on 23rd Jan 16:23:03 <scuttlemonkey> cool beans 16:23:11 <lalatenduM> yes 16:23:45 <lalatenduM> anything to discuss? else time to close the meeting :0) 16:24:09 <lalatenduM> 1 16:24:11 <lalatenduM> 2 16:24:13 <lalatenduM> 3 16:24:18 <scuttlemonkey> sold! 16:24:18 <lalatenduM> #endmeeting