| *** tstclair has quit IRC | 00:23 | |
| *** kbroughton has joined #savanna | 00:28 | |
| *** tstclair has joined #savanna | 00:40 | |
| *** IlyaE has joined #savanna | 02:16 | |
| *** antigluk has quit IRC | 02:17 | |
| *** SergeyLukjanov has joined #savanna | 03:06 | |
| *** SergeyLukjanov has quit IRC | 04:17 | |
| *** IlyaE has quit IRC | 06:23 | |
| *** bogdando has joined #savanna | 07:19 | |
| *** SergeyLukjanov has joined #savanna | 07:32 | |
| openstackgerrit | Sergey Reshetnyak proposed a change to stackforge/savanna: Fix Cinder volumes support with xenserver https://review.openstack.org/47792 | 08:14 |
|---|---|---|
| openstackgerrit | Sergey Reshetnyak proposed a change to stackforge/savanna: Fix Cinder volumes support with xenserver https://review.openstack.org/47792 | 08:20 |
| *** nprivalova has joined #savanna | 08:34 | |
| *** ruhe has joined #savanna | 09:29 | |
| openstackgerrit | A change was merged to stackforge/savanna-extra: Remove version pbr pins from setup_requires https://review.openstack.org/47731 | 09:35 |
| openstackgerrit | A change was merged to stackforge/python-savannaclient: Sync requiremnts with global requirements https://review.openstack.org/47725 | 09:35 |
| openstackgerrit | A change was merged to stackforge/savanna-dashboard: Remove version pbr pins from setup_requires https://review.openstack.org/47727 | 09:35 |
| openstackgerrit | A change was merged to stackforge/python-savannaclient: Remove version pbr pins from setup_requires https://review.openstack.org/47726 | 09:35 |
| openstackgerrit | Alexander Ignatov proposed a change to stackforge/savanna: Doc fix for replacement of Hadoop version in Vanilla plugin https://review.openstack.org/47378 | 09:37 |
| openstackgerrit | A change was merged to stackforge/savanna: Impl context.to_dict() https://review.openstack.org/47751 | 09:39 |
| openstackgerrit | A change was merged to stackforge/python-savannaclient: Adjusting json resource names https://review.openstack.org/47587 | 09:39 |
| openstackgerrit | A change was merged to stackforge/savanna-dashboard: Sync requirements with global requirements https://review.openstack.org/47728 | 09:40 |
| openstackgerrit | A change was merged to stackforge/savanna: Remove version pbr pins from setup_requires https://review.openstack.org/47722 | 09:40 |
| openstackgerrit | A change was merged to stackforge/savanna: Sync requirements with global requirements https://review.openstack.org/47723 | 09:40 |
| *** ruhe has quit IRC | 10:05 | |
| *** ruhe has joined #savanna | 10:11 | |
| *** ruhe has quit IRC | 10:12 | |
| *** nprivalova has quit IRC | 10:12 | |
| *** nprivalova has joined #savanna | 10:29 | |
| *** SergeyLukjanov has quit IRC | 10:36 | |
| *** dmitryme has joined #savanna | 10:41 | |
| *** ruhe has joined #savanna | 10:50 | |
| *** SergeyLukjanov has joined #savanna | 11:05 | |
| openstackgerrit | Nikolay Mahotkin proposed a change to stackforge/savanna: Add /job-configs/<job_type> REST API call https://review.openstack.org/45419 | 11:06 |
| *** _crobertsrh is now known as crobertsrh | 11:08 | |
| openstackgerrit | A change was merged to stackforge/savanna: Add default sqlite db to .gitignore https://review.openstack.org/47752 | 11:18 |
| openstackgerrit | Sergey Reshetnyak proposed a change to stackforge/savanna: Fix Cinder volumes support with xenserver https://review.openstack.org/47792 | 11:26 |
| *** ruhe has quit IRC | 11:27 | |
| openstackgerrit | Nikita Konovalov proposed a change to stackforge/savanna-dashboard: Fixed scaling cluster message. https://review.openstack.org/47817 | 11:40 |
| openstackgerrit | Nikita Konovalov proposed a change to stackforge/savanna-dashboard: Fixed scaling cluster message https://review.openstack.org/47817 | 11:43 |
| openstackgerrit | Nadya Privalova proposed a change to stackforge/savanna-dashboard: JobOrigin UI impl https://review.openstack.org/46910 | 11:45 |
| *** ruhe has joined #savanna | 11:46 | |
| openstackgerrit | Nikolay Mahotkin proposed a change to stackforge/savanna: Add /job-configs/<job_type> REST API call https://review.openstack.org/45419 | 11:49 |
| *** dmitryme has quit IRC | 11:59 | |
| openstackgerrit | A change was merged to stackforge/savanna-dashboard: Fixed scaling cluster message https://review.openstack.org/47817 | 12:00 |
| openstackgerrit | Vadim Rovachev proposed a change to stackforge/savanna-dashboard: Added UI tests https://review.openstack.org/43442 | 12:27 |
| *** dmitryme has joined #savanna | 12:33 | |
| openstackgerrit | Andrew Lazarev proposed a change to stackforge/savanna-extra: Implemented data locality for Swift https://review.openstack.org/47824 | 12:37 |
| openstackgerrit | Nikita Konovalov proposed a change to stackforge/savanna-dashboard: Status bar added for Job Execution https://review.openstack.org/47825 | 12:39 |
| *** ruhe has quit IRC | 12:50 | |
| openstackgerrit | Nikita Konovalov proposed a change to stackforge/savanna-dashboard: Status bar added for Job Execution https://review.openstack.org/47825 | 12:56 |
| openstackgerrit | Nikolay Mahotkin proposed a change to stackforge/savanna: Fix submitting hive job https://review.openstack.org/47000 | 12:58 |
| openstackgerrit | Nikolay Mahotkin proposed a change to stackforge/savanna: Fix submitting hive job https://review.openstack.org/47000 | 13:02 |
| *** kbroughton has quit IRC | 13:06 | |
| mattf | nprivalova, seen http://paste.fedoraproject.org/41536/13799420/ before? | 13:15 |
| dmitryme | mattf, we just discussed that with other folks. We didn't see that before | 13:24 |
| dmitryme | I would suggest looking at 'nova_network.cidr' via debug output | 13:24 |
| mattf | i think it's because i set user_neutron=true in config instead of use_neutron=true | 13:24 |
| mattf | the instances were coming up attached to all networks and savanna was hitting the init_nova instead of init_neutron code path | 13:25 |
| dmitryme | yep, sounds very probable | 13:26 |
| *** kbroughton has joined #savanna | 13:28 | |
| *** SergeyLukjanov has quit IRC | 13:29 | |
| *** SergeyLukjanov has joined #savanna | 13:34 | |
| mattf | dmitryme, so i'm trying out the neutron support | 13:34 |
| mattf | i've an external network and an admin network | 13:34 |
| mattf | savanna-api is on the controller | 13:34 |
| *** tmckay has joined #savanna | 13:34 | |
| mattf | i start a cluster w/ the "management network" equal to my admin-network | 13:35 |
| *** kbroughton_ has joined #savanna | 13:35 | |
| mattf | the admin-network is private, ext-net is public and can get floating ips | 13:35 |
| mattf | as you'd expect, savanna-api can't connect to any of the instances, they're on a separate netns | 13:35 |
| mattf | i can ip netns exec ssh to get to instances | 13:35 |
| *** NikitaKonovalov has joined #savanna | 13:35 | |
| *** kbroughton has quit IRC | 13:35 | |
| *** kbroughton_ is now known as kbroughton | 13:35 | |
| mattf | dmitryme, any suggestion on how i should tweak my config? | 13:36 |
| dmitryme | could you decipher "i can ip netns exec ssh to get to instances"? | 13:37 |
| dmitryme | ok, Nikita told me what that means | 13:39 |
| mattf | from the controller, i can run "ip netns exec ssh root@nodeip" -- let's me run ssh w/i the instance's netns | 13:39 |
| dmitryme | can't you use ext-net as Savanna management network? | 13:40 |
| mattf | that will give each instance an external ip -- i have a limited number of those | 13:41 |
| mattf | i'm trying it now though | 13:41 |
| NikitaKonovalov | mattf, Savanna can run only pure ssh commands, so management_ip should be reachable without netns | 13:44 |
| mattf | NikitaKonovalov, what's your suggestion for how to ensure savanna can reach the management ip? | 13:46 |
| *** NikitaKonovalov has quit IRC | 13:56 | |
| mattf | dmitryme, anyone else over there have a suggestion on how to make sure savanna-api can reach the mgmt ip? | 14:01 |
| dmitryme | I assume you can't use IPs from ext-net as management, right? | 14:03 |
| mattf | i couldn't get dhcp working there, the instances would come up but would not get their ip | 14:04 |
| dmitryme | actually Savanna can assign floating ip from pool | 14:04 |
| dmitryme | so you don't need dhcp | 14:04 |
| mattf | savanna managing that until neutron supports auto-assign i assume | 14:05 |
| mattf | what's the config for that? | 14:05 |
| dmitryme | ops, it is not merged | 14:06 |
| dmitryme | https://review.openstack.org/#/c/44822/ | 14:06 |
| dmitryme | could you use that patch set directly for now | 14:07 |
| dmitryme | ? | 14:07 |
| dmitryme | until we merge it | 14:07 |
| dmitryme | I think we'll do it pretty soon | 14:07 |
| dmitryme | corresponding UI change: | 14:07 |
| dmitryme | https://review.openstack.org/#/c/45245/ | 14:07 |
| SergeyLukjanov | yep, I think that we'll wait for additional reviews one more day and than merge it | 14:08 |
| mattf | yeah, i was afraid that'd be the one | 14:08 |
| mattf | ultimately we want to be able to support a deployment where savanna-api is on the controllers, the head of the cluster gets an externally routable ip (either floating or making head instance multi-homed) and the workers of the cluster are off in a private netns | 14:09 |
| mattf | agreed on ^^ ? | 14:10 |
| SergeyLukjanov | yep, but it looks like that it'll require netns support or better architecture | 14:12 |
| dmitryme | with the patch set above merged, Savanna will be able to do that. But only if Savanna can reach instances private IPs | 14:12 |
| mattf | i think the key bit is "savanna can reach instance private IPs", which is independent of the floating ip patch | 14:12 |
| SergeyLukjanov | dmitryme, we discussed earlier that we can potentially spawn ssh processes with netns? any updates on it? | 14:13 |
| dmitryme | am, I think you've discussed that with somebody else :-) | 14:13 |
| mattf | SergeyLukjanov, we discussed putting a proxy agent in the master node before too | 14:14 |
| SergeyLukjanov | mattf, the same problem could be with nova-network too, and, so, we have an use-floating-ips flag for it | 14:14 |
| mattf | savanna-api on control -> savanna-api (or like) on master, then have the master config the rest of the cluter | 14:14 |
| SergeyLukjanov | mattf, there are some problems with agents too - how agent will communicate and auth on controller, etc. | 14:15 |
| *** rnirmal has joined #savanna | 14:15 | |
| mattf | that solution moves the overhead of config onto the cluster itself, frees up savanna-api on controller | 14:15 |
| SergeyLukjanov | it could be solved by splitting savanna to savanna-api and savanna-engine (that could be deployed on computes for example) | 14:16 |
| mattf | SergeyLukjanov, re comm w/ auth, that's where the discussion stopped between us. i don't follow why the agent will need access to anything outside the cluster | 14:16 |
| SergeyLukjanov | mattf, I mean access to savanna-api | 14:16 |
| mattf | savanna-api on the controller should poll the savanna-agent on the master | 14:17 |
| SergeyLukjanov | mattf, it's the same thing like ssh to the hadoop cluster's node | 14:17 |
| mattf | i claim taht polling is cheaper than ssh'ing and config'ing each node in the cluster | 14:17 |
| SergeyLukjanov | but it will not solve problem with accessibility using private IPs | 14:18 |
| dmitryme | obviously that is true, but for big clusters we can't use just a single master node for configuring all others | 14:18 |
| dmitryme | that will take too much time | 14:18 |
| mattf | in any event, we do have a hurdle here and two options. 0) netns enable savanna-api 1) move config into an agent. (0) is faster and more inline with what we have now, (1) will pop up again when savanna-api is provisioning large clusters | 14:18 |
| mattf | SergeyLukjanov, (1) will solve the private ip problem, because the master will be on the private network | 14:19 |
| SergeyLukjanov | 0) will be required for polling agents too | 14:19 |
| *** ruhe has joined #savanna | 14:19 | |
| dmitryme | mattf, in agent design, how savanna-api will access master node, via public IP? | 14:20 |
| mattf | dmitryme, in all cases the master node needs to be on two networks, the public network and the cluster network. right now those two networks have to be the same. | 14:20 |
| *** NikitaKonovalov has joined #savanna | 14:20 | |
| SergeyLukjanov | I would like to clarify that I think that agents are potentially good solution for some problems, but it's very complex update of architecture and it should be discussed at the design summit | 14:21 |
| mattf | in all cases savanna-api has to have access to the master node. in (1) savanna-api ONLY has to have access to the master node. | 14:21 |
| mattf | SergeyLukjanov, yeah, (1) is non-sense for 0.3, (0) is a candidate for 0.3 | 14:22 |
| SergeyLukjanov | mattf, oh, got it, it's about the problem about we can't assign public IPs to all 1000 nodes in big clusters | 14:22 |
| SergeyLukjanov | mattf, additionally we should remember about Heat integration for orchestration | 14:23 |
| dmitryme | yep, Heat is a good point | 14:23 |
| mattf | savanna-api <-> {public} <-> [head] <-> {private} <-> [worker]* | 14:23 |
| mattf | anyone know how that'd work if you s/savanna-api/heat/ ? | 14:23 |
| dmitryme | we didn't investigate it yet | 14:24 |
| ruhe | use floating ip | 14:24 |
| dmitryme | but possibly we will still have to do some stuff via SSH | 14:24 |
| dmitryme | even with Heat utilized | 14:24 |
| SergeyLukjanov | yeah, possible | 14:25 |
| SergeyLukjanov | dmitryme, I hope that we could use netns for ssh in current arch too | 14:25 |
| mattf | so state of the art is neutron works only if you can put instances on the public network (i've had issues with this), once the previiously mentioned patch is applied savanna will auto-assign floating ips, in the future we still have to handle the separate cluster network issue | 14:26 |
| mattf | for handling separate cluster network we have netns enablement and agent arch as two solutions | 14:26 |
| mattf | ^^ we on the same page? | 14:27 |
| dmitryme | sounds like a good summary | 14:27 |
| ruhe | agree, requirement for direct access to VMs is an architectural problems. | 14:27 |
| mattf | ok, ttyl, otp for me | 14:27 |
| *** NikitaKonovalov has quit IRC | 14:30 | |
| *** NikitaKonovalov has joined #savanna | 14:34 | |
| *** nprivalova has quit IRC | 14:49 | |
| openstackgerrit | A change was merged to stackforge/savanna-dashboard: Floating IP assignement support https://review.openstack.org/45245 | 14:57 |
| openstackgerrit | A change was merged to stackforge/savanna-dashboard: Status bar added for Job Execution https://review.openstack.org/47825 | 15:13 |
| openstackgerrit | A change was merged to stackforge/savanna: Fix Cinder volumes support with xenserver https://review.openstack.org/47792 | 15:14 |
| openstackgerrit | A change was merged to stackforge/savanna: Doc fix for replacement of Hadoop version in Vanilla plugin https://review.openstack.org/47378 | 15:18 |
| openstackgerrit | A change was merged to stackforge/savanna: Set default log levels for some third-party libs https://review.openstack.org/47740 | 15:23 |
| openstackgerrit | A change was merged to stackforge/savanna: Temporarily fixes bug #1223934 https://review.openstack.org/47724 | 15:26 |
| *** IlyaE has joined #savanna | 15:30 | |
| *** sacharya has joined #savanna | 15:37 | |
| openstackgerrit | Ilya Tyaptin proposed a change to stackforge/savanna: Trusts for longrunning tasks https://review.openstack.org/44706 | 15:40 |
| *** nprivalova has joined #savanna | 15:42 | |
| *** IlyaE has quit IRC | 15:53 | |
| *** NikitaKonovalov has quit IRC | 15:53 | |
| openstackgerrit | Sergey Lukjanov proposed a change to stackforge/savanna-dashboard: JobOrigin UI impl https://review.openstack.org/46910 | 16:10 |
| openstackgerrit | A change was merged to stackforge/savanna-dashboard: JobOrigin UI impl https://review.openstack.org/46910 | 16:11 |
| *** IlyaE has joined #savanna | 16:31 | |
| *** mattf is now known as _mattf | 16:45 | |
| *** SergeyLukjanov has quit IRC | 16:56 | |
| tmckay | bah | 17:08 |
| tmckay | fyi, I finally got an instance to run on F19. The last ultimate error was this: | 17:08 |
| tmckay | sudo sysctl -w fs.protected_hardlinks=0 | 17:08 |
| tmckay | apparently on F19, you have to turn hard link protection off on root owned files or libguestfs breaks. I found it buried in a BZ. | 17:09 |
| tmckay | now, maybe, I can actually run a cluster... | 17:09 |
| *** dmitryme has quit IRC | 17:26 | |
| *** SergeyLukjanov has joined #savanna | 17:31 | |
| *** dmitryme has joined #savanna | 17:36 | |
| *** ruhe has quit IRC | 17:38 | |
| nprivalova | tmckay, crobertsrh, ping | 18:01 |
| tmckay | hi | 18:02 |
| nprivalova | today we saw a demo with EDP UI | 18:02 |
| tmckay | wow, great | 18:03 |
| nprivalova | and found several problems... | 18:03 |
| tmckay | oh | 18:03 |
| nprivalova | :) | 18:03 |
| *** ruhe has joined #savanna | 18:03 | |
| *** ruhe has quit IRC | 18:03 | |
| nprivalova | fields input-type, output-type looks unreasonable in Job | 18:04 |
| nprivalova | field job-type it's better to be set in jobOrigin. because the set of binaries determines type of the job | 18:05 |
| nprivalova | so…we may skip Job origin creation at all …. | 18:05 |
| nprivalova | we need your feedback very much | 18:06 |
| crobertsrh | Hmm, that's an interesting idea | 18:07 |
| tmckay | do you mean not have a Job Origin object, or create it automatically? | 18:07 |
| nprivalova | no, just remove Job | 18:08 |
| nprivalova | we need JobOrigin | 18:08 |
| *** dmitryme has quit IRC | 18:09 | |
| nprivalova | job_type ==> JobOrigin, input_type, output_type are not needed because we have the type in DataSource | 18:09 |
| nprivalova | all configs may be set during execution | 18:10 |
| tmckay | so a job_execution points straight to a job_origin | 18:12 |
| nprivalova | yes | 18:12 |
| tmckay | the origin represents the list of scripts, and a job_execution represents an instance (with configs) | 18:13 |
| tmckay | makes sense to me | 18:13 |
| nprivalova | crobertsrh, what's your thoughts? | 18:14 |
| crobertsrh | might be a more concise way to do things | 18:15 |
| tmckay | crobertsrh, do you mean "there might be an even more concise way to do things" or "this might be a more concise way to do things than what we have now" | 18:15 |
| openstackgerrit | Ilya Tyaptin proposed a change to stackforge/savanna: Trusts for longrunning tasks https://review.openstack.org/44706 | 18:16 |
| crobertsrh | the later | 18:16 |
| tmckay | thanks | 18:16 |
| nprivalova | ok, guys, I think I will prepare CR for tomorrow with changes | 18:16 |
| tmckay | hurray! :) | 18:17 |
| nprivalova | crobertsrh, just skip all bugs about Job page (if there are some) | 18:18 |
| crobertsrh | ok, sounds good. Great idea | 18:18 |
| *** ruhe has joined #savanna | 18:51 | |
| *** ruhe has quit IRC | 18:51 | |
| *** dmitryme has joined #savanna | 18:52 | |
| *** nprivalova has quit IRC | 18:54 | |
| *** dmitryme has quit IRC | 19:05 | |
| *** _mattf is now known as mattf | 19:21 | |
| *** mattf is now known as _mattf | 19:25 | |
| *** ruhe has joined #savanna | 19:52 | |
| *** akuznetsov has joined #savanna | 19:55 | |
| *** akuznetsov has quit IRC | 20:03 | |
| *** jfriedly has quit IRC | 20:09 | |
| *** jfriedly has joined #savanna | 20:11 | |
| *** ruhe has quit IRC | 20:19 | |
| *** SergeyLukjanov has quit IRC | 20:30 | |
| *** tmckay has quit IRC | 21:19 | |
| *** _mattf is now known as mattf | 22:04 | |
| *** IlyaE has quit IRC | 22:14 | |
| *** sacharya has quit IRC | 22:25 | |
| *** openstack has joined #savanna | 22:32 | |
| *** IlyaE has joined #savanna | 22:35 | |
| *** kbroughton has quit IRC | 23:03 | |
| *** rnirmal has quit IRC | 23:15 | |
| *** matsuhashi has joined #savanna | 23:50 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!