| *** littleidea has quit IRC | 00:07 | |
| *** utlemming has quit IRC | 00:07 | |
| *** rbasak has quit IRC | 00:08 | |
| *** flaviamissi has joined #openstack-dev | 00:09 | |
| *** mnewby has quit IRC | 00:10 | |
| *** anotherjesse is now known as anotherjesse_zz | 00:11 | |
| *** andrewsben_zz is now known as andrewsben | 00:17 | |
| *** utlemming has joined #openstack-dev | 00:19 | |
| *** vladimir3p has quit IRC | 00:22 | |
| *** roge has quit IRC | 00:25 | |
| *** jgriffith has quit IRC | 00:27 | |
| *** jgriffith has joined #openstack-dev | 00:27 | |
| *** jakedahn is now known as jakedahn_zz | 00:30 | |
| *** maplebed has quit IRC | 00:30 | |
| *** edygarcia has joined #openstack-dev | 00:38 | |
| *** s0mik has quit IRC | 00:44 | |
| *** rods has quit IRC | 00:46 | |
| *** flaviamissi_ has joined #openstack-dev | 00:46 | |
| *** flaviamissi has quit IRC | 00:47 | |
| *** flaviamissi_ is now known as flaviamissi | 00:47 | |
| *** vipul has quit IRC | 00:49 | |
| *** issackelly has quit IRC | 00:50 | |
| *** mdomsch has quit IRC | 00:52 | |
| *** aa has quit IRC | 00:55 | |
| *** pixelbeat has quit IRC | 00:58 | |
| *** reed has quit IRC | 01:02 | |
| *** ayoung has joined #openstack-dev | 01:03 | |
| *** maoy has joined #openstack-dev | 01:14 | |
| *** thingee has quit IRC | 01:16 | |
| *** longhudou has quit IRC | 01:18 | |
| *** sdake has joined #openstack-dev | 01:18 | |
| *** jdurgin has quit IRC | 01:19 | |
| *** thingee has joined #openstack-dev | 01:19 | |
| *** ijw has quit IRC | 01:19 | |
| *** heyho has quit IRC | 01:19 | |
| *** thingee has quit IRC | 01:19 | |
| *** dweimer has quit IRC | 01:20 | |
| *** zaitcev has quit IRC | 01:30 | |
| *** andrewsben is now known as andrewsben_zz | 01:36 | |
| *** zaitcev has joined #openstack-dev | 01:37 | |
| *** dweimer has joined #openstack-dev | 01:40 | |
| *** anotherjesse_zz is now known as anotherjesse | 01:43 | |
| *** adalbas has quit IRC | 01:46 | |
| *** wiliam has quit IRC | 01:50 | |
| *** adjohn has quit IRC | 01:54 | |
| *** anotherjesse is now known as anotherjesse_zz | 01:58 | |
| *** lorin1 has joined #openstack-dev | 02:00 | |
| *** danwent has quit IRC | 02:00 | |
| lorin1 | Is the "[no]foo" boolean flag format still kosher with ini format? Or is it just "foo=false" these days? | 02:00 |
|---|---|---|
| *** jeblair has quit IRC | 02:01 | |
| *** wiliam has joined #openstack-dev | 02:04 | |
| *** andrewsben_zz is now known as andrewsben | 02:05 | |
| *** reed has joined #openstack-dev | 02:09 | |
| *** epim has joined #openstack-dev | 02:14 | |
| *** andrewsben is now known as andrewsben_zz | 02:16 | |
| *** timello_ has quit IRC | 02:26 | |
| *** timello has joined #openstack-dev | 02:37 | |
| *** jeblair has joined #openstack-dev | 02:40 | |
| *** bhuvan_ has quit IRC | 02:48 | |
| *** s0mik has joined #openstack-dev | 02:49 | |
| *** gabrielhurley has quit IRC | 02:50 | |
| *** sandywalsh_ has quit IRC | 02:52 | |
| *** littleidea has joined #openstack-dev | 02:54 | |
| *** somik has joined #openstack-dev | 02:54 | |
| *** jgriffith has quit IRC | 02:54 | |
| *** s0mik has quit IRC | 02:57 | |
| *** somik is now known as s0mik | 02:57 | |
| *** timello has quit IRC | 03:01 | |
| *** jakedahn_zz is now known as jakedahn | 03:06 | |
| *** edygarcia has quit IRC | 03:14 | |
| *** jakedahn is now known as jakedahn_zz | 03:15 | |
| *** jakedahn_zz is now known as jakedahn | 03:16 | |
| *** johnpostlethwait has joined #openstack-dev | 03:16 | |
| *** edygarcia has joined #openstack-dev | 03:17 | |
| *** lorin1 has quit IRC | 03:21 | |
| *** koolhead17 has quit IRC | 03:25 | |
| *** ywu has quit IRC | 03:32 | |
| *** Ryan_Lane has quit IRC | 03:36 | |
| *** edygarcia has quit IRC | 03:37 | |
| *** jgriffith has joined #openstack-dev | 03:40 | |
| *** johnpostlethwait has quit IRC | 03:41 | |
| *** s0mik has quit IRC | 03:50 | |
| *** issackel_ has joined #openstack-dev | 03:55 | |
| *** issackel_ has quit IRC | 03:56 | |
| *** s0mik has joined #openstack-dev | 04:02 | |
| *** wiliam has quit IRC | 04:11 | |
| *** s0mik has quit IRC | 04:15 | |
| *** danwent has joined #openstack-dev | 04:15 | |
| *** anderstj has joined #openstack-dev | 04:20 | |
| *** anotherjesse_zz is now known as anotherjesse | 04:29 | |
| *** harlowja has quit IRC | 04:29 | |
| *** anotherjesse is now known as anotherjesse_zz | 04:30 | |
| *** anotherjesse_zz is now known as anotherjesse | 04:30 | |
| *** jgriffith has quit IRC | 04:31 | |
| *** sleepsonzzz is now known as sleepsonthefloor | 04:31 | |
| *** danwent has quit IRC | 04:47 | |
| vishy | lorin1: foo=false | 04:49 |
| *** vipul has joined #openstack-dev | 05:05 | |
| *** anderstj has quit IRC | 05:05 | |
| *** sleepsonthefloor is now known as sleepsonzzz | 05:10 | |
| *** maoy has quit IRC | 05:10 | |
| *** s0mik has joined #openstack-dev | 05:13 | |
| *** reed has quit IRC | 05:33 | |
| *** utlemming has quit IRC | 05:35 | |
| *** Ryan_Lane has joined #openstack-dev | 05:38 | |
| *** Ryan_Lane has joined #openstack-dev | 05:38 | |
| *** zaitcev has quit IRC | 05:39 | |
| *** sleepsonzzz is now known as sleepsonthefloor | 05:43 | |
| zykes- | vishy: I've done it | 05:44 |
| zykes- | but it still starts with qemu | 05:45 |
| vishy | zykes-: I don't believe you :) | 05:45 |
| vishy | does nova-compute mention libvirt_type=kvm when it lists flags on startup? | 05:45 |
| zykes- | nope | 05:46 |
| vishy | then you haven't set the fag properly | 05:47 |
| vishy | :) | 05:47 |
| vishy | * flag | 05:47 |
| zykes- | I'm sorry I mean it does | 05:47 |
| zykes- | libvirt_type : kvm from < nova-compute -d | 05:48 |
| zykes- | why does ubuntu use nova-compute.conf and nova.conf ? | 05:49 |
| davidha | Anyone still awake? | 05:49 |
| vishy | zykes-: sigh no se | 05:50 |
| davidha | :) Does swift return HTTP headers indicating how long an objct can be cached? | 05:50 |
| vishy | zykes-: but if you launch a new instance it still doesn't start with kvm? | 05:50 |
| vishy | zykes-: does the libvirt.xml still specify qemu? | 05:50 |
| zykes- | need to get to work before I can try again vishy :) | 05:51 |
| vishy | zykes-: my guess is that you missed something :) | 05:51 |
| zykes- | maybe :) | 05:51 |
| *** anotherjesse is now known as anotherjesse_zz | 05:54 | |
| *** anderstj has joined #openstack-dev | 05:57 | |
| *** danwent has joined #openstack-dev | 06:01 | |
| *** hattwick has quit IRC | 06:01 | |
| *** dubsquared has quit IRC | 06:03 | |
| *** anderstj has quit IRC | 06:10 | |
| *** dachary has joined #openstack-dev | 06:11 | |
| *** mindpixel has joined #openstack-dev | 06:13 | |
| *** dachary has quit IRC | 06:16 | |
| *** s0mik has quit IRC | 06:23 | |
| *** rgoodwin has quit IRC | 06:24 | |
| *** jakedahn is now known as jakedahn_zz | 06:24 | |
| *** vincentricci has quit IRC | 06:25 | |
| *** jakedahn_zz is now known as jakedahn | 06:27 | |
| *** rgoodwin has joined #openstack-dev | 06:33 | |
| zykes- | vishy: do you know why ubuntu has 2 nova compute files ? | 06:40 |
| zykes- | as in /etc/nova/nova-compute.conf and nova.conf | 06:40 |
| *** mrunge has joined #openstack-dev | 06:44 | |
| zykes- | so it was all the nova-compute.conf that made issues | 06:44 |
| *** pmezard has joined #openstack-dev | 06:45 | |
| *** vipul has quit IRC | 06:46 | |
| *** garyk has quit IRC | 06:46 | |
| *** davidha has quit IRC | 06:49 | |
| *** vipul has joined #openstack-dev | 06:52 | |
| *** sleepsonthefloor is now known as sleepsonzzz | 07:01 | |
| *** journeeman has joined #openstack-dev | 07:07 | |
| *** reidrac has joined #openstack-dev | 07:07 | |
| *** dachary has joined #openstack-dev | 07:17 | |
| *** garyk has joined #openstack-dev | 07:21 | |
| zykes- | jk0: here for some help on openstack-bosh-cpi | 07:26 |
| *** Stackops-Jorge has joined #openstack-dev | 07:26 | |
| *** Stackops-Jorge has quit IRC | 07:31 | |
| *** davidha has joined #openstack-dev | 07:31 | |
| *** Stackops-Jorge has joined #openstack-dev | 07:33 | |
| *** vipul has quit IRC | 07:38 | |
| *** dtheodor has joined #openstack-dev | 07:57 | |
| *** Adri2000_ has joined #openstack-dev | 07:58 | |
| *** Adri2000_ has quit IRC | 07:58 | |
| *** Adri2000_ has joined #openstack-dev | 07:58 | |
| *** hattwick has joined #openstack-dev | 07:58 | |
| *** apevec has joined #openstack-dev | 07:58 | |
| *** derekh has joined #openstack-dev | 07:59 | |
| *** Adri2000 has quit IRC | 08:03 | |
| *** littleidea has quit IRC | 08:19 | |
| *** darraghb has joined #openstack-dev | 08:21 | |
| *** dingdeng has quit IRC | 08:22 | |
| *** dingdeng has joined #openstack-dev | 08:22 | |
| *** pmezard has quit IRC | 08:32 | |
| *** pmezard has joined #openstack-dev | 08:35 | |
| *** pixelbeat has joined #openstack-dev | 08:36 | |
| *** gael has joined #openstack-dev | 08:37 | |
| *** zigo has joined #openstack-dev | 08:45 | |
| *** zigo-_- has joined #openstack-dev | 08:50 | |
| *** zigo has quit IRC | 08:50 | |
| *** danpb has joined #openstack-dev | 09:02 | |
| *** mcclurmc_ has quit IRC | 09:16 | |
| ttx | koolhead17 almost scored the millionth bug in LP: | 09:30 |
| ttx | bug 999999 | 09:30 |
| uvirtbot | Launchpad bug 999999 in keystone "keystone does not support postgres database" [Undecided,New] https://launchpad.net/bugs/999999 | 09:30 |
| Daviey | so close... but yet, no dice. | 09:31 |
| *** davidha_who_took has joined #openstack-dev | 09:44 | |
| *** davidha has quit IRC | 09:44 | |
| *** littleidea has joined #openstack-dev | 09:55 | |
| *** lts has joined #openstack-dev | 10:10 | |
| *** mancdaz has joined #openstack-dev | 10:12 | |
| *** heyho has joined #openstack-dev | 10:22 | |
| *** pvo-away is now known as pvo | 10:25 | |
| *** littleidea has quit IRC | 10:25 | |
| *** ijw has joined #openstack-dev | 10:30 | |
| *** davidha has joined #openstack-dev | 10:37 | |
| *** davidha_who_took has quit IRC | 10:38 | |
| zykes- | jk0: could you provide a sample .yml file for micro bosh ? | 10:43 |
| *** eglynn has joined #openstack-dev | 10:58 | |
| *** eglynn__ has quit IRC | 11:00 | |
| *** epim_ has joined #openstack-dev | 11:02 | |
| *** epim has quit IRC | 11:06 | |
| *** epim_ is now known as epim | 11:06 | |
| *** flaviamissi has quit IRC | 11:14 | |
| *** rkukura has quit IRC | 11:17 | |
| *** wiliam has joined #openstack-dev | 11:29 | |
| *** markvoelker has joined #openstack-dev | 11:34 | |
| *** Adri2000_ is now known as Adri2000 | 11:35 | |
| *** Mkenneth has joined #openstack-dev | 11:37 | |
| *** dhellmann has quit IRC | 11:44 | |
| *** troytoman-away is now known as troytoman | 11:45 | |
| *** salgado has joined #openstack-dev | 11:46 | |
| *** salgado has joined #openstack-dev | 11:46 | |
| LarsErikP | if i want to file a bug on the documentation.. where do i put that? | 11:54 |
| *** mcclurmc_ has joined #openstack-dev | 11:55 | |
| LarsErikP | nvm.. found it | 11:59 |
| *** ayoung has quit IRC | 12:03 | |
| *** bsza has joined #openstack-dev | 12:07 | |
| *** adalbas has joined #openstack-dev | 12:12 | |
| *** dprince has joined #openstack-dev | 12:17 | |
| *** cdub has joined #openstack-dev | 12:25 | |
| *** rbasak has joined #openstack-dev | 12:25 | |
| *** rbasak has joined #openstack-dev | 12:25 | |
| *** roge has joined #openstack-dev | 12:25 | |
| *** mrunge has quit IRC | 12:25 | |
| *** maoy has joined #openstack-dev | 12:28 | |
| *** mikal has quit IRC | 12:35 | |
| *** cdub has quit IRC | 12:37 | |
| *** mikal has joined #openstack-dev | 12:37 | |
| *** tryggvil has joined #openstack-dev | 12:40 | |
| *** alaski has joined #openstack-dev | 12:41 | |
| *** openstackjenkins has quit IRC | 12:42 | |
| *** openstackjenkins has joined #openstack-dev | 12:43 | |
| *** tryggvil has quit IRC | 12:45 | |
| *** sandywalsh has joined #openstack-dev | 12:45 | |
| *** flaviamissi has joined #openstack-dev | 12:45 | |
| *** mancdaz has quit IRC | 12:52 | |
| *** mancdaz has joined #openstack-dev | 12:53 | |
| *** openstackjenkins has quit IRC | 12:55 | |
| *** openstackjenkins has joined #openstack-dev | 12:55 | |
| *** mestery has quit IRC | 12:56 | |
| *** stuntmachine has joined #openstack-dev | 12:56 | |
| *** mestery has joined #openstack-dev | 12:57 | |
| *** openstackjenkins has quit IRC | 12:59 | |
| zykes- | notmyname: ping pong | 12:59 |
| zykes- | does swift support regional replication now ? | 13:00 |
| *** openstackjenkins has joined #openstack-dev | 13:00 | |
| *** openstackjenkins has quit IRC | 13:01 | |
| *** openstackjenkins has joined #openstack-dev | 13:02 | |
| *** lorin1 has joined #openstack-dev | 13:02 | |
| *** lorin1 has left #openstack-dev | 13:03 | |
| *** dolphm_ has joined #openstack-dev | 13:03 | |
| *** openstackjenkins has quit IRC | 13:03 | |
| *** openstackjenkins has joined #openstack-dev | 13:03 | |
| notmyname | zykes-: no change in status for that feature. (status == probably work to some degree, but untested and there are probably problems that need to be solved) | 13:03 |
| *** openstackjenkins has joined #openstack-dev | 13:04 | |
| *** openstackjenkins has quit IRC | 13:10 | |
| *** openstackjenkins has joined #openstack-dev | 13:10 | |
| *** wiliam has quit IRC | 13:11 | |
| *** Mkenneth has quit IRC | 13:12 | |
| *** alaski has quit IRC | 13:12 | |
| *** alaski has joined #openstack-dev | 13:13 | |
| *** sandywalsh_ has joined #openstack-dev | 13:22 | |
| *** sandywalsh has quit IRC | 13:22 | |
| *** zaitcev has joined #openstack-dev | 13:24 | |
| *** wiliam has joined #openstack-dev | 13:24 | |
| *** ayoung has joined #openstack-dev | 13:27 | |
| *** mancdaz has quit IRC | 13:28 | |
| *** mancdaz has joined #openstack-dev | 13:29 | |
| dolphm_ | ayoung: pki question when you have some time :) | 13:31 |
| ayoung | dolphm_, sure | 13:31 |
| dolphm_ | ayoung: so, i was thinking this morning about the backwards compatibility of adding pki -- is that something you've thought about? | 13:32 |
| ayoung | yes | 13:32 |
| dolphm_ | ayoung: are your thoughts documented somewhere? | 13:32 |
| ayoung | dolphm_, yes, just still editing them prior to public posting | 13:32 |
| ayoung | but the short is | 13:32 |
| openstackgerrit | Verification of a change to openstack/nova failed: Sync with latest version of openstack.common.cfg https://review.openstack.org/7483 | 13:32 |
| ayoung | it should be OK | 13:33 |
| ayoung | modulo the token ID being in the URL | 13:33 |
| ayoung | but that is problematic for other reasons | 13:33 |
| dolphm_ | ayoung: length? | 13:33 |
| ayoung | dolphm_, I'll show you | 13:33 |
| *** adalbas has quit IRC | 13:34 | |
| ayoung | dolphm_, this is a Base64 encoded signed document http://fpaste.org/T7qg/ | 13:34 |
| ayoung | so while that will fit into -X-Auth_header | 13:34 |
| ayoung | doing http://keystonehost/v2.0/tokens/<that thing} is prohibitive | 13:34 |
| dolphm_ | ayoung: yeah, that's starting to get risky | 13:35 |
| ayoung | dolphm_, the thing is, there is already an understanding that putting the token into the URL is bad security | 13:36 |
| dolphm_ | ayoung: are you familiar with the X-Subject-Token approach from ~e3? | 13:36 |
| ayoung | dolphm_, no | 13:36 |
| dolphm_ | ayoung: it was a solution to the token-in-url issue that didn't make it into the keystone rewrite | 13:36 |
| ayoung | dolphm_, link? | 13:36 |
| *** Mkenneth has joined #openstack-dev | 13:37 | |
| *** timello has joined #openstack-dev | 13:37 | |
| dolphm_ | ayoung: i'll try and find one | 13:37 |
| dolphm_ | https://blueprints.launchpad.net/keystone/+spec/stop-ids-in-uris | 13:37 |
| ayoung | dolphm_, so...assuming that there is an alternative to the token-in-Url scheme (which I assume there will be) then the PKI poken scheme is just a really long token | 13:37 |
| ayoung | dolphm_, ah, yes, I've seen that | 13:38 |
| dolphm_ | ayoung: the idea is that auth still occurs via X-Auth-Token, and instead of passing a token ID in the URL, you'd pass in the "subject" of your request as an X-Subject-Token header | 13:38 |
| dolphm_ | ayoung: i was never really happy with the resulting resources (they're not very restful) | 13:38 |
| *** markmcclain has joined #openstack-dev | 13:39 | |
| dolphm_ | ayoung: i liked the idea of splitting a "token" into two values: an ID (which is insecure and safe to use in the URL) and a value, which is what needs to be protected, and passed exclusively in headers | 13:39 |
| ayoung | dolphm_, we can still do that. | 13:40 |
| ayoung | the Base64 text would be the value. | 13:40 |
| dolphm_ | so, for a service to validate a token, they'd do the regular GET /tokens/{token_id} (token_id being the insecure identifier) along with an X-Subject-Token or X-Token-Value header or something with the actual token value to be validated | 13:41 |
| ayoung | So what ever is done in base Keystone for tokens will be done for PKI tokens. Thus, I feel pretty confident about backwards compat | 13:41 |
| dolphm_ | plus the service's own X-Auth-Token | 13:41 |
| dolphm_ | ayoung: could a pki token be handled blindly by the API as unencrypted? | 13:41 |
| ayoung | dolphm_, yep | 13:41 |
| *** openstackjenkins has quit IRC | 13:41 | |
| dolphm_ | ayoung: awesome, that's what i was hoping for | 13:41 |
| ayoung | its is just a longer token | 13:42 |
| *** openstackjenkins has joined #openstack-dev | 13:42 | |
| *** dhellmann has joined #openstack-dev | 13:42 | |
| * dolphm_ keeps getting more and more excited for pki! | 13:42 | |
| ayoung | I'll have proof of concept code posted soonish | 13:42 |
| ayoung | I can show what needs to be done from the command line using the NSS utils | 13:43 |
| ayoung | I'm just trying to make is legible | 13:43 |
| *** rkukura has joined #openstack-dev | 13:44 | |
| *** maoy has quit IRC | 13:45 | |
| *** openstackjenkins has quit IRC | 13:47 | |
| *** openstackjenkins has joined #openstack-dev | 13:48 | |
| *** apevec has quit IRC | 13:49 | |
| *** openstackjenkins has quit IRC | 13:49 | |
| *** lorin1 has joined #openstack-dev | 13:50 | |
| *** openstackjenkins has joined #openstack-dev | 13:50 | |
| *** lorin1 has left #openstack-dev | 13:50 | |
| *** joesavak has joined #openstack-dev | 13:51 | |
| *** PotHix has joined #openstack-dev | 13:53 | |
| *** edygarcia has joined #openstack-dev | 13:55 | |
| *** adalbas has joined #openstack-dev | 13:58 | |
| *** crobinso has joined #openstack-dev | 14:04 | |
| *** salgado has quit IRC | 14:04 | |
| ayoung | dolphm_, I'm splitting the article into two parts for clarity. Here;s the first part. http://adam.younglogic.com/2012/05/signing-certutil/ | 14:06 |
| *** salgado has joined #openstack-dev | 14:07 | |
| *** salgado has joined #openstack-dev | 14:07 | |
| dolphm_ | ayoung: thanks, i'll read it this morning | 14:07 |
| zykes- | jk0: ping | 14:09 |
| *** danwent has quit IRC | 14:12 | |
| *** journeeman has quit IRC | 14:12 | |
| *** b1rkh0ff has joined #openstack-dev | 14:13 | |
| *** dtroyer is now known as dtroyer_zzz | 14:13 | |
| *** b1rkh0ff has quit IRC | 14:14 | |
| *** b1rkh0ff has joined #openstack-dev | 14:15 | |
| *** b1rkh0ff has quit IRC | 14:15 | |
| *** b1rkh0ff has joined #openstack-dev | 14:15 | |
| *** sleepsonzzz is now known as sleepsonthefloor | 14:17 | |
| *** jsavak has joined #openstack-dev | 14:20 | |
| *** andrewsben_zz is now known as andrewsben | 14:21 | |
| *** joesavak has quit IRC | 14:23 | |
| *** GheRivero has joined #openstack-dev | 14:24 | |
| *** openstackjenkins has quit IRC | 14:26 | |
| davidha | notmyname, hi | 14:26 |
| notmyname | davidha: good morning | 14:26 |
| *** openstackjenkins has joined #openstack-dev | 14:26 | |
| *** dtroyer_zzz is now known as dtroyer | 14:27 | |
| davidha | notmyname, gm, is there any support to controlling the HTTP headers indicating how long an object may be cached? | 14:27 |
| davidha | notmyname: I mean when a client does get-object - will the http headers indicating how long this object can be cached be set? | 14:29 |
| notmyname | davidha: I think the answer is "yes", but the question doesn't really make sense to me. this is for caching in an external system | 14:29 |
| notmyname | ? | 14:29 |
| davidha | yes - for external cache | 14:29 |
| davidha | (one object may be cacheable while a second non cacheable) | 14:30 |
| notmyname | davidha:ok, that makes sense. I was a little worried you were thinking about some sort of cache inside of swift (which doesn't exist) | 14:30 |
| davidha | notmyname: S3 seem to have this feature - hence my Q :) | 14:30 |
| notmyname | davidha: for cloudfront? | 14:30 |
| notmyname | davidha: you can add any piece of arbitrary metadata to any object. either use the "x-object-meta-key: value" form or add your own header to the allowed_headers config variable in the object server config and then set that | 14:32 |
| davidha | notmyname: Well I do not know if it part of cloudFront or not. | 14:35 |
| notmyname | davidha: you have a caching layer in front of swift and want to store the TTL with the object, right? | 14:35 |
| davidha | But you answer is that swift is extendiable to incldue any header per object | 14:35 |
| davidha | notmyname: yes - as an example | 14:36 |
| davidha | you=your | 14:36 |
| notmyname | davidha: ya, so add whatever you want the header to be named to the object server config and then start setting it on your objects | 14:37 |
| notmyname | (or let your customers start setting it) | 14:37 |
| davidha | notmyname: I get it. | 14:37 |
| *** tserong has quit IRC | 14:39 | |
| ayoung | dolphm_, posted the full proof of concept now, too: http://adam.younglogic.com/2012/05/signing-certutil/ | 14:39 |
| *** danwent has joined #openstack-dev | 14:39 | |
| davidha | creiht: gm. | 14:39 |
| *** sdake has quit IRC | 14:39 | |
| creiht | morning | 14:40 |
| *** sdake has joined #openstack-dev | 14:40 | |
| davidha | Thanks for your comments yesterday. I do not think it all boils down to concurency - concurency can be bad - depending on what limits the request processor. | 14:41 |
| creiht | well I was more talking about the limitations of python in general in terms of concurrency | 14:42 |
| *** sleepsonthefloor is now known as sleepsonzzz | 14:43 | |
| *** maoy has joined #openstack-dev | 14:44 | |
| *** renier has quit IRC | 14:45 | |
| *** rnirmal has joined #openstack-dev | 14:48 | |
| *** openstackjenkins has quit IRC | 14:48 | |
| davidha | ok, at the long run, I would like to see the wsgi web front end as an option for getting a server running using a minimalistic web front end while actual deployments can choose their best server (i.e. not necesserily fastest server, but best overall). | 14:48 |
| *** openstackjenkins has joined #openstack-dev | 14:48 | |
| davidha | creiht: (the eventlet wsgi, being one option they can choose from) | 14:49 |
| *** blamar has joined #openstack-dev | 14:49 | |
| *** anderstj has joined #openstack-dev | 14:52 | |
| *** anderstj has quit IRC | 14:55 | |
| *** matwood_ has joined #openstack-dev | 14:56 | |
| *** mindpixel has quit IRC | 14:59 | |
| *** matwood has quit IRC | 14:59 | |
| *** openstackjenkins has quit IRC | 15:00 | |
| *** matwood_ has quit IRC | 15:00 | |
| *** openstackjenkins has joined #openstack-dev | 15:00 | |
| creiht | davidha: as long as that can be done simply/generically I would be fine with that | 15:00 |
| davidha | creiht: great - I think it can be quiet simple for Swift - no idea about other servers | 15:02 |
| davidha | I will cont working to get it working. | 15:02 |
| notmyname | davidha: to echo what creiht said, I like your idea in general. unfortunately, it's hard to say if it's better or not until you actually have it working. but it's certainly stuff I'd consider including in swift | 15:03 |
| davidha | quiet=quite | 15:03 |
| *** maoy has quit IRC | 15:05 | |
| *** maoy has joined #openstack-dev | 15:05 | |
| davidha | notmyname - sounds good. If it will be better depends on what you measure with. Its likely that it would be better for some in the community while others be better off with the existing solution. It will make swift more open for sure. Lets play this and see where it goes. | 15:09 |
| *** troytoman is now known as troytoman-away | 15:12 | |
| *** garyk has quit IRC | 15:16 | |
| *** garyk has joined #openstack-dev | 15:16 | |
| *** blamar has quit IRC | 15:17 | |
| *** utlemming has joined #openstack-dev | 15:18 | |
| *** Ryan_Lane has quit IRC | 15:18 | |
| *** vladimir3p has joined #openstack-dev | 15:19 | |
| *** openstackjenkins has quit IRC | 15:22 | |
| *** openstackjenkins has joined #openstack-dev | 15:23 | |
| *** matwood has joined #openstack-dev | 15:24 | |
| *** matwood has quit IRC | 15:24 | |
| *** dtheodor has quit IRC | 15:25 | |
| s1rp | vishy: dprince: i pushed up a new version of https://review.openstack.org/#/c/7458/ ; included a docstring explaining why per-table utf8 is needed for now | 15:26 |
| *** openstackjenkins has quit IRC | 15:27 | |
| *** kbringard has joined #openstack-dev | 15:27 | |
| *** jgriffith has joined #openstack-dev | 15:27 | |
| *** openstackjenkins has joined #openstack-dev | 15:27 | |
| *** matwood has joined #openstack-dev | 15:28 | |
| *** cdub has joined #openstack-dev | 15:29 | |
| *** mancdaz has quit IRC | 15:36 | |
| *** mancdaz has joined #openstack-dev | 15:38 | |
| *** reidrac has quit IRC | 15:40 | |
| *** lloydde has joined #openstack-dev | 15:41 | |
| *** devananda has quit IRC | 15:42 | |
| *** devananda has joined #openstack-dev | 15:43 | |
| *** vladimir3p has quit IRC | 15:44 | |
| *** jgriffith has quit IRC | 15:46 | |
| *** jgriffith has joined #openstack-dev | 15:47 | |
| *** rnirmal_ has joined #openstack-dev | 15:47 | |
| *** rnirmal has quit IRC | 15:48 | |
| *** rnirmal_ is now known as rnirmal | 15:48 | |
| *** rnirmal has quit IRC | 15:50 | |
| *** rnirmal has joined #openstack-dev | 15:51 | |
| *** matwood has quit IRC | 15:51 | |
| *** matwood has joined #openstack-dev | 15:51 | |
| *** simo has joined #openstack-dev | 15:56 | |
| *** clayg has quit IRC | 15:56 | |
| *** clayg_ is now known as clayg | 15:56 | |
| openstackgerrit | Verification of a change to openstack/nova failed: Add version to compute rpc API. https://review.openstack.org/7391 | 15:57 |
| *** clayg_ has joined #openstack-dev | 15:57 | |
| ayoung | simo, no I still don't think revocations are necessary | 15:58 |
| ayoung | just that if we have them, they could be pushed out from Keystone on a queue | 15:58 |
| andrewbogott_ | dhellmann: I'm adding some commands to openstackclient, and one of them is producing "ERROR cliff.app PrettyTable instance has no attribute 'align' " | 15:59 |
| ayoung | TTL in the queue based on the expiriy of the tokens they are revoking | 15:59 |
| andrewbogott_ | dhellmann: My questions are 1) Any guess what's causing that? and 2) Would it be possible for things like that to be exceptions rather than log messages so that we get a stack trace? | 15:59 |
| simo | ayoung: tehy are not absolutely necessary, but they make administration easier, as you can revoke a token when you change roles for a user | 15:59 |
| dhellmann | andrewbogott_, the version of PrettyTable used by the other client apps is older and has a slightly different API than what cliff wants | 15:59 |
| dhellmann | use the --debug option to get a stack trace | 16:00 |
| dhellmann | openstackclient --debug your command here | 16:00 |
| andrewbogott_ | Oh, nice! | 16:00 |
| ayoung | simo, right now, the CLI gets tokens before evey call, it does not even cache them. To say that we need to expire tokens would be a major increase in the lifespan of a token | 16:00 |
| simo | ayoung: will you allow any service to ask a token-cert ? Or do you want to require proof of auth before giving out a token ? | 16:00 |
| andrewbogott_ | dhellmann: OK, so I need to update PrettyTable? Or will that break all the other clients? | 16:00 |
| dhellmann | we should probably do something to trap openstack exceptions but let the others through by default | 16:00 |
| ayoung | proof-of-auth | 16:00 |
| simo | ayoung: proof-of-auth means services cannot ask for a token if they are close to expiration time | 16:01 |
| dhellmann | I updated via "sudo pip install -U PrettyTable" and both "nova list" and "openstack list server" work on the same system, but that was surprising (I expected the nova command to break) | 16:01 |
| simo | ayoung: how do you handle potential failures due to expired tokens ? | 16:01 |
| ayoung | simo, lets call them signed-tokens. As they will not be certificates in any X509 meaning of the term when we get to the final impl | 16:01 |
| dhellmann | you might want to install openstackclient into a virtualenv? | 16:01 |
| simo | or do you need at all ? | 16:01 |
| ayoung | simo, resubmit | 16:01 |
| simo | ayoung: authz-token then | 16:01 |
| simo | ayoung: resubmit means the job is killed and the admin needs to resubmit it ? | 16:02 |
| ayoung | simo, no, signed tokens, to distinguish them from the current impl, which I would call cookies | 16:02 |
| ayoung | simo, yes, as all of this is authenticated upon jub submission | 16:02 |
| simo | signed-authz-token (just teasing) | 16:02 |
| simo | ayoung: ok but a job may take long time and then try to do an action | 16:03 |
| simo | if the job always take > expiration time then it will always fail on the final action | 16:03 |
| simo | making the job impossible to complete | 16:03 |
| simo | ayoung: is there any case like that ? | 16:03 |
| simo | I am thinking something like provisiona machine then do an action and provisioning turns out to take 1 full hour | 16:04 |
| andrewbogott_ | dhellmann: Worked! thanks. | 16:04 |
| dhellmann | my guess is nova does not try to align columns | 16:05 |
| dtroyer | dhellmann, andrewbogott_: the other clients all handle pretty table 0.6 correctly now. devstack is still pinned to 0.5, I think that can be removed now | 16:05 |
| dhellmann | ah, cool | 16:05 |
| ayoung | simo, when I submit a job, say through nova, the token tells nova that I am authed to submit it. Then nova does a bunch of stuff on my behalf, and probably using the data from the token. But Nova has its own long lived auth via an admin token. Now, admin tokens I could see an argument for revocation. | 16:05 |
| ayoung | But your argument could be extended to HPC type jobs that take months. THere is always the possibility of a process that will outlast its auth token. | 16:07 |
| simo | ayoung: wait you mean user auth-tokens are not carried on to authorize all steps ? | 16:07 |
| simo | ayoung: yes there is always the possibility in kerberos we have ways to solve the problem, I am just saying you may need those ways too | 16:07 |
| simo | ayoung: for example allowing the release of a new token, bound to a service. by presenting evidence of a non-expired token | 16:08 |
| ayoung | eglynn, when nova talks to glance on behalf of a user, which token is used to authorize the action? THe users token, or the admin token? | 16:08 |
| *** troytoman-away is now known as troytoman | 16:09 | |
| eglynn | ayoung: I'd expect the user's token is propogated to glance by nova | 16:09 |
| eglynn | ayoung: (otherwise every user effectively becomes an admin, right?) | 16:10 |
| ayoung | eglynn, depends. Admin API can say "do this on behalf of user X" | 16:11 |
| ayoung | But I don't know glance well enough to say whether that would make sense. | 16:11 |
| ayoung | simo, OK, so currently tokens are just cookies. Which means that their lifespan is effectively controlled by Keystone. To time out a cookie-token, the server just responds unauthorized | 16:12 |
| eglynn | ayoung: I don't know if the simple x-auth-token header style of creds propogation supports OBO-style operation | 16:12 |
| *** mancdaz_ has joined #openstack-dev | 16:13 | |
| eglynn | ayoung: or even, if it could be done within the current scheme, whether it would really bring anything to the table | 16:13 |
| ayoung | eglynn, so you would expect nova to pass on the token it was presented. Thus, those tokens have to live long enough to perform the whole operation. | 16:13 |
| eglynn | ayoung: yep (default token expiry is 24hrs IIRC) | 16:14 |
| *** Gordonz has joined #openstack-dev | 16:14 | |
| *** Lumiere has quit IRC | 16:15 | |
| *** smoser has joined #openstack-dev | 16:15 | |
| eglynn | ayoung: 24hrs unless of course the token is revoked in the meantime (derekh recently proposed a revocation mechanism for cases where a passwd was compromised) | 16:15 |
| *** Gordonz has quit IRC | 16:15 | |
| ayoung | simo, edygarcia so now roles are checked on line, which means that if I revoke a role at the 12 hour mark, future operations can use that same token, and will just fail if the user no longer has the role required | 16:15 |
| *** edygarcia has quit IRC | 16:16 | |
| *** Gordonz has joined #openstack-dev | 16:16 | |
| ayoung | eglynn, ^^ was for you, of course | 16:16 |
| *** mancdaz has quit IRC | 16:16 | |
| *** mancdaz_ is now known as mancdaz | 16:16 | |
| *** anotherjesse_zz is now known as anotherjesse | 16:16 | |
| *** Lumiere has joined #openstack-dev | 16:17 | |
| smoser | ok. so i'm clearly reading something wrong. but doesn't _mounted_processing in nova/virt/xenapi/vm_utils.py invoke inject_data_into_fs from nova.virt.disk incorrectly ? | 16:17 |
| simo | eglynn: how is revocation performed ? | 16:17 |
| smoser | it looks to me like its passing a varible named 'metadata' into a variable named 'admin_passwd' while passing in 'None' for 'metadata' | 16:17 |
| eglynn | ayoung: one sec ... | 16:18 |
| jgriffith | anotherjesse: ping | 16:18 |
| ayoung | derekh, is your "Revocation" proposal written up anywhere> | 16:18 |
| *** sleepsonzzz is now known as sleepsonthefloor | 16:18 | |
| ayoung | sleepsonthefloor, https://bugzilla.redhat.com/show_bug.cgi?id=822182 and https://bugzilla.redhat.com/show_bug.cgi?id=822187 | 16:19 |
| uvirtbot | bugzilla.redhat.com bug 822182 in Package Review "Review Request: python-websockify - Python proxy for the websockets protocol" [Medium,New] | 16:19 |
| anotherjesse | jgriffith pong - sending a direct message | 16:19 |
| eglynn | ayoung see https://review.openstack.org/7276 & https://review.openstack.org/7344 | 16:20 |
| *** Mkenneth has quit IRC | 16:20 | |
| *** Mkenneth1 has joined #openstack-dev | 16:20 | |
| eglynn | ayoung: ^^^ wrt token revocation | 16:20 |
| sleepsonthefloor | ayoung: looking | 16:21 |
| ayoung | eglynn, OK. Thanks. simo I concede the point, and will write up the revocation scheme. I'll include both the HTTP and AMQP based options | 16:21 |
| *** salgado has quit IRC | 16:22 | |
| *** maplebed has joined #openstack-dev | 16:24 | |
| *** edygarcia has joined #openstack-dev | 16:24 | |
| *** salgado has joined #openstack-dev | 16:25 | |
| *** salgado has joined #openstack-dev | 16:25 | |
| *** reed has joined #openstack-dev | 16:27 | |
| simo | ayoung: cool | 16:27 |
| derekh | ayoung: https://review.openstack.org/#/c/7276/ | 16:27 |
| *** s0mik has joined #openstack-dev | 16:28 | |
| ayoung | derekh, thanks | 16:28 |
| derekh | ayoung: and there is another for when a user is disabled https://review.openstack.org/#/c/7344/ | 16:28 |
| ayoung | derekh, thanks...I see that one is already merged | 16:32 |
| derekh | ayoung: yup one to invalidate tokens when a users password changes merged the other is still waiting | 16:33 |
| ayoung | derekh, what about if a user is removed from a tenancy? I guess it is OK to leave the token around, and just return "No roles" when they online check is made? | 16:34 |
| derekh | ayoung: good question, I havn't looked into that case | 16:35 |
| *** littleidea has joined #openstack-dev | 16:36 | |
| *** gael has quit IRC | 16:37 | |
| ayoung | derekh, I'm trying to address the needs for revocation WRT to the PKI approach to token validation that I wrote up here http://adam.younglogic.com/2012/05/signed-authz-authn/ | 16:37 |
| *** vipul has joined #openstack-dev | 16:38 | |
| *** danwent has quit IRC | 16:38 | |
| *** issackel_ has joined #openstack-dev | 16:39 | |
| *** zigo-_- has quit IRC | 16:40 | |
| *** zigo has joined #openstack-dev | 16:40 | |
| *** garyk has quit IRC | 16:41 | |
| *** andrewsben is now known as andrewsben_zz | 16:41 | |
| *** sdake has quit IRC | 16:42 | |
| derekh | ayoung: I'll take a look through it now | 16:42 |
| ayoung | derekh, thanks | 16:42 |
| *** AlanClark has joined #openstack-dev | 16:44 | |
| bhuvan | can someone review (and approve) https://review.openstack.org/7149 and https://review.openstack.org/7175? | 16:47 |
| bcwaldon | Vek: ping | 16:47 |
| bhuvan | fwiw, 7149 is for python-keystoneclient and 7175 is for keystone | 16:48 |
| *** danwent has joined #openstack-dev | 16:49 | |
| *** rgoodwin is now known as rgoodwin_away | 16:49 | |
| openstackgerrit | Verification of a change to openstack/keystone failed: Changed arguments in keystone CLI for consistency. https://review.openstack.org/7253 | 16:51 |
| *** adjohn has joined #openstack-dev | 16:52 | |
| *** mnewby has joined #openstack-dev | 16:53 | |
| *** vipul has quit IRC | 17:01 | |
| *** andresambrois has joined #openstack-dev | 17:02 | |
| *** zigo has quit IRC | 17:03 | |
| *** zigo has joined #openstack-dev | 17:03 | |
| *** mcclurmc_ has quit IRC | 17:08 | |
| derekh | ayoung: I had a look through your post at first glance to me the idea makes sense, I gotta run but I'll try out your examples later | 17:09 |
| ayoung | derekh, thanks for the check...I would say sanity check, but I would not want to make unsubstantiated claims about myself | 17:10 |
| derekh | ayoung: :-) | 17:10 |
| *** derekh has quit IRC | 17:11 | |
| *** maoy has quit IRC | 17:15 | |
| *** danpb has quit IRC | 17:15 | |
| *** _adjohn has joined #openstack-dev | 17:16 | |
| *** Blah1 has joined #openstack-dev | 17:17 | |
| *** rgoodwin_away is now known as rgoodwin | 17:19 | |
| *** adjohn has quit IRC | 17:19 | |
| *** _adjohn is now known as adjohn | 17:19 | |
| *** Blah1 has quit IRC | 17:20 | |
| *** harlowja has joined #openstack-dev | 17:20 | |
| *** Stackops-Jorge has quit IRC | 17:25 | |
| *** vincentricci has joined #openstack-dev | 17:27 | |
| *** Blah1 has joined #openstack-dev | 17:28 | |
| *** Ryan_Lane has joined #openstack-dev | 17:28 | |
| *** Ryan_Lane has joined #openstack-dev | 17:28 | |
| *** GheRivero has quit IRC | 17:32 | |
| comstud | russellb: left an idea on your rpc versioning review.. ping me if you want to talk here or brainstorm some more. | 17:34 |
| *** davidha has quit IRC | 17:36 | |
| *** anderstj has joined #openstack-dev | 17:38 | |
| mtaylor | danwent, mnewby: https://review.openstack.org/#/c/7494/ - should fix tarball jobs | 17:41 |
| mtaylor | bcwaldon, jk0: https://review.openstack.org/#/c/7490/ should fix tarball jobs | 17:41 |
| zykes- | jk0: ping | 17:41 |
| *** matwood has quit IRC | 17:42 | |
| danwent | mtaylor: what testing (if any) is needed on this by reviews? | 17:43 |
| *** anderstj_ has joined #openstack-dev | 17:43 | |
| mtaylor | danwent: shouldn't need to be any ... it should pass the gate - but I did run it locally to make sure | 17:43 |
| mtaylor | danwent: it shouldn't change normal behavior | 17:43 |
| *** anderstj has quit IRC | 17:43 | |
| danwent | mtaylor: ok, i figured, but just confirming | 17:43 |
| russellb | comstud: cool, looking over it now. I actually didn't use RpcDispatcher as a mix-in on the simpler conversions. When I got to scheduler and compute, I did, because it was easier ... I like not using it as a mix-in better, though. | 17:46 |
| comstud | ahh..i only really reviewed the compute versioning review | 17:47 |
| comstud | and so it there | 17:47 |
| comstud | so far | 17:47 |
| comstud | so it there/saw it there | 17:47 |
| * russellb nods | 17:47 | |
| russellb | your patch may make it easier to not use it as a mix-in | 17:48 |
| russellb | which would rock | 17:48 |
| *** _0x44 has quit IRC | 17:49 | |
| comstud | yeah, it makes your versioning changes easier, actually | 17:49 |
| comstud | because it'll default all managers to 1.0 | 17:49 |
| comstud | since I modified the base class | 17:49 |
| comstud | and you won't have to mix it in manually | 17:49 |
| comstud | though I didn't go back and double check your changes to the managers to make sure that statement is truly accurate | 17:49 |
| comstud | hehe | 17:49 |
| *** _0x44 has joined #openstack-dev | 17:50 | |
| *** jdurgin has joined #openstack-dev | 17:50 | |
| comstud | but essentially.. I'd like a more generic way to be able to create a dispatcher | 17:50 |
| russellb | so if a manager wanted to have more than one api version served up, would it just override create_rpc_dispatcher() ? | 17:50 |
| comstud | and i don't necessarily like having to set the .managers attribute manually... | 17:50 |
| russellb | (in your patch) | 17:51 |
| comstud | niceer to be able to be to pass them to __init__ | 17:51 |
| comstud | correct | 17:51 |
| russellb | yeah, that is nicer indeed | 17:51 |
| comstud | you could switch part of it back to using API_VERSION.. | 17:51 |
| comstud | i'm not really opposed to that | 17:51 |
| *** mrunge has joined #openstack-dev | 17:51 | |
| *** andrewsben_zz is now known as andrewsben | 17:52 | |
| comstud | I had it as API_VERSION = '1.0' in the base manager class at first | 17:52 |
| russellb | instead of the tuple? | 17:52 |
| comstud | yea | 17:52 |
| russellb | k | 17:52 |
| comstud | i don't necessarily like the tuple... but.. i don't necessarily like the API_VERSION getattr magic, either. haha | 17:52 |
| comstud | but it's less important to me | 17:52 |
| russellb | i kind of like having it there ... in your face right by the code | 17:52 |
| comstud | than being able to pass a list of managers or callbacks to RpcDispatcher | 17:53 |
| russellb | important thing is not forgetting to update it when it should be updated | 17:53 |
| *** littleidea has quit IRC | 17:53 | |
| comstud | nod | 17:53 |
| russellb | cool, thanks much for the feedback | 17:53 |
| comstud | sure.. this also simplifies the rpc code and makes it always use the dispatcher now | 17:53 |
| russellb | i shold be able to incorporate all this stuff this afternoon | 17:53 |
| russellb | yeah, i like that | 17:53 |
| *** johnpostlethwait has joined #openstack-dev | 17:54 | |
| comstud | sorry it took me a while to get to reviewing | 17:54 |
| comstud | def would love to see this get in asap | 17:54 |
| russellb | it didn't take that long | 17:54 |
| *** Ryan_Lane has quit IRC | 17:54 | |
| comstud | well, i thought you'd put this up last week or something | 17:54 |
| comstud | :) | 17:54 |
| russellb | yeah, these patches love to go into conflict | 17:54 |
| russellb | nah, Monday I think | 17:54 |
| comstud | ahh ok | 17:54 |
| comstud | been heads down lately on some internal stuff @ rax | 17:55 |
| russellb | i had it on github before that, you may have seen me mention that on the ML | 17:55 |
| *** darraghb has quit IRC | 17:55 | |
| * russellb nods | 17:55 | |
| comstud | ahh that might be it | 17:55 |
| russellb | payin' billz | 17:55 |
| comstud | ya :) | 17:55 |
| comstud | finally updated my cells branch | 17:55 |
| comstud | need to get the BP created | 17:56 |
| russellb | nice | 17:56 |
| russellb | and document it a bit? :-p | 17:56 |
| comstud | haha | 17:56 |
| comstud | that's the idea | 17:56 |
| russellb | maybe you have, i dunno, haven't looked | 17:56 |
| russellb | though i'm defnitely interested in that work | 17:56 |
| mtaylor | bcwaldon: oops. I forgot a git add. thanks. | 17:57 |
| comstud | if you incorporate my changes, I'll be able to make the code make a little more sense | 17:57 |
| comstud | cuz I will separate out some of the worker methods into different classes | 17:57 |
| russellb | cool | 17:57 |
| comstud | kinda nasty right now | 17:57 |
| russellb | gotta start somewhere | 17:58 |
| russellb | that's a big (complex) problem to solve | 17:58 |
| russellb | so i have another blueprint that got assigned to me ... formalized-message-structures | 17:59 |
| russellb | i was thinking that by roping all rpc usage into nova/*/rpcapi.py modules, that sort of accomplishes the goal | 17:59 |
| russellb | that make sense, or do you think there is more that should be done? | 18:00 |
| comstud | I like the rpcapi modules | 18:01 |
| mtaylor | mnewby: fixed that patch ... thanks for pointing out that I forgot to remove stuff from setup.cf | 18:01 |
| mtaylor | cfg | 18:01 |
| comstud | i like centralizing the creation of the messages | 18:01 |
| comstud | instead of forming the rpc messages all over the places | 18:02 |
| russellb | cool, glad you like it ... it was time consuming :-p | 18:02 |
| mnewby | mtaylor: np | 18:02 |
| comstud | Yeah | 18:02 |
| *** jog0 has joined #openstack-dev | 18:02 | |
| comstud | I'm not familiar with that blueprint, but sounds like it should address at least part of it | 18:02 |
| russellb | Documenting the messages passed through the rpc layer to improve modularity and scalability. | 18:02 |
| russellb | that's the whole description, heh | 18:02 |
| comstud | ahhh | 18:03 |
| comstud | then yeah, blueprint done. | 18:03 |
| russellb | score | 18:03 |
| openstackgerrit | Verification of a change to openstack/keystone failed: Bug 890411: service roles and ID filtering https://review.openstack.org/7010 | 18:03 |
| comstud | Just link docs to rpcapi.pys ;) | 18:03 |
| uvirtbot | Launchpad bug 890411 in keystone "Tenant role conflicts/overlaps can be a security issue" [Medium,In progress] https://launchpad.net/bugs/890411 | 18:03 |
| *** somik has joined #openstack-dev | 18:03 | |
| russellb | comstud: thanks again for the feedback. i'm off to incorporate your changes | 18:03 |
| comstud | no prob | 18:04 |
| *** adjohn has quit IRC | 18:05 | |
| *** s0mik has quit IRC | 18:06 | |
| *** somik is now known as s0mik | 18:06 | |
| *** epim_ has joined #openstack-dev | 18:11 | |
| anotherjesse | jgriffith: building a dependency tree for cinder transition here: https://blueprints.launchpad.net/nova/+spec/remove-nova-volumes | 18:13 |
| anotherjesse | just getting started | 18:13 |
| russellb | comstud: oh, hrm, you want more than one API class for the same API version. That will require another tweak to the dispatch() method but I think that's ok | 18:14 |
| russellb | comstud: that's what you meant by breaking up your cells code? | 18:14 |
| *** zigo has quit IRC | 18:14 | |
| *** epim has quit IRC | 18:14 | |
| *** epim_ is now known as epim | 18:14 | |
| comstud | russellb: They don't have to be versioned the same | 18:15 |
| comstud | russellb: In fact, I think they shouldn't be.. | 18:15 |
| comstud | that's why I have the tuples | 18:15 |
| comstud | but | 18:15 |
| comstud | what you had before by using manager.API_VERSION would also still work | 18:15 |
| russellb | right now it stops on the first is_compatible() | 18:16 |
| comstud | not sure if I understood your version correctly | 18:16 |
| *** jog0 has left #openstack-dev | 18:16 | |
| comstud | I think that's fine.. | 18:16 |
| comstud | one shouldn't define the same method in multiple classes.. | 18:16 |
| comstud | maybe | 18:17 |
| comstud | :) | 18:17 |
| russellb | unless it's 2 API versions | 18:17 |
| russellb | this example you put on gerrit ... return rpc_dispatcher.RpcDispatcher([(RpcCallSetX(), '1.0'), (RpcCallSetY(), '1.0')]) | 18:17 |
| comstud | yeah... maybe then you pick the closest version | 18:17 |
| comstud | however, i wasn't thinking about that case | 18:17 |
| russellb | right now it would never call anything on RpcCallSetY() | 18:17 |
| comstud | my goal was just to split up functionality | 18:17 |
| *** mnewby has quit IRC | 18:17 | |
| comstud | a set of rpc methods for cell scheduler | 18:17 |
| comstud | a set of rpc methods for cell routing | 18:17 |
| comstud | different method names tho | 18:18 |
| russellb | k, so i just need to make it ... if is_compatible() and hasattr(): | 18:18 |
| comstud | just makes up the huge manager class | 18:18 |
| russellb | right | 18:18 |
| comstud | into some smaller classes | 18:18 |
| comstud | oooh | 18:18 |
| comstud | yeah | 18:18 |
| comstud | I missed that | 18:18 |
| comstud | what was the goal with allowing multiple managers in the first place? | 18:18 |
| russellb | easy change, just making sure that's what we want | 18:18 |
| comstud | just curious | 18:18 |
| russellb | supporting a 1.0 and 2.0 at the same time | 18:19 |
| russellb | (eventually) | 18:19 |
| russellb | if we're like, this API sucks, i want to rewrite it (but support messages from the old version for upgrades) | 18:19 |
| *** asisin has quit IRC | 18:19 | |
| comstud | ooh right | 18:19 |
| comstud | duh | 18:20 |
| *** davidha has joined #openstack-dev | 18:20 | |
| russellb | normally we'd just be doing little changes, and bumping it to 1.1, 1.2, etc | 18:20 |
| russellb | meaning it's compatible with 1.X where X <= the current version | 18:21 |
| russellb | or so was my thinking | 18:21 |
| *** gyee has joined #openstack-dev | 18:25 | |
| *** alaski has quit IRC | 18:26 | |
| *** davidha has quit IRC | 18:27 | |
| *** novas0x2a|laptop has joined #openstack-dev | 18:27 | |
| *** davidha has joined #openstack-dev | 18:27 | |
| *** alaski has joined #openstack-dev | 18:28 | |
| *** maoy has joined #openstack-dev | 18:30 | |
| *** heyho has quit IRC | 18:30 | |
| maoy | vishy: ping | 18:30 |
| *** jakedahn is now known as jakedahn_zz | 18:30 | |
| *** mnewby has joined #openstack-dev | 18:31 | |
| *** mnewby has joined #openstack-dev | 18:32 | |
| *** dolphm_ has quit IRC | 18:32 | |
| *** smoser has quit IRC | 18:32 | |
| *** wting has quit IRC | 18:33 | |
| *** smoser has joined #openstack-dev | 18:33 | |
| *** bourke has quit IRC | 18:33 | |
| *** wting has joined #openstack-dev | 18:34 | |
| *** rkukura has quit IRC | 18:34 | |
| maoy | comstud: ping | 18:35 |
| cp16net | jgriffith: hey i added the notifications :) https://review.openstack.org/7517 | 18:35 |
| *** bourke has joined #openstack-dev | 18:35 | |
| jgriffith | cp16net: Awesome!!! I'll go have a look | 18:36 |
| comstud | russellb: makes sense! | 18:38 |
| comstud | maoy: pong | 18:38 |
| maoy | comstud: https://docs.google.com/document/d/1nlKmYld3xxpTv6Xx0Iky6L46smbEqg7-SWPu_o6VJws/edit | 18:39 |
| maoy | comstud: i'm trying to see if there is a way to better describe the nova state machine and simplify it | 18:39 |
| *** adjohn has joined #openstack-dev | 18:40 | |
| maoy | comstud: i'd like to get some early feedback before i send to the ML | 18:40 |
| *** adjohn has quit IRC | 18:41 | |
| *** adjohn has joined #openstack-dev | 18:41 | |
| comstud | there's a larger problem I think I'd like considered along with this... | 18:41 |
| maoy | comstud: listening | 18:41 |
| maoy | comstud: it's part of a larger problem i'm working on too. :) | 18:41 |
| comstud | in that the API should keep track of its own view of the state | 18:42 |
| *** alaski has quit IRC | 18:42 | |
| comstud | and updates should be pushed to it when needed.. | 18:42 |
| *** alaski has joined #openstack-dev | 18:42 | |
| comstud | so.. | 18:42 |
| comstud | what API needs and what compute worker needs might be completely different | 18:43 |
| comstud | it's almost like task_state is what the worker cares about | 18:43 |
| comstud | and vm_state is more what the API cares about | 18:43 |
| comstud | although we do have an extension to expose task_state as well via API | 18:43 |
| comstud | but anyway.. :) | 18:43 |
| maoy | comstud: API should care task_state too to avoid running concurrent conflicting tasks | 18:43 |
| *** jsavak has quit IRC | 18:44 | |
| comstud | as best as it can, anyway | 18:44 |
| comstud | the problem is that API should always treat the information as at least slightly stale. | 18:44 |
| *** garyk has joined #openstack-dev | 18:44 | |
| comstud | this is important for a highly distributed setup | 18:45 |
| comstud | esp w/ cells | 18:45 |
| comstud | because w/ cells, API has its own completely separate table | 18:45 |
| comstud | you can have a cell in AU with compute workers updating state.. | 18:46 |
| maoy | i haven't looked at the cell stuff at all. | 18:46 |
| comstud | but say you have an API server somewhere that currently can't talk to it | 18:46 |
| comstud | certainly they're not using the same DB.. | 18:46 |
| comstud | so updates from that cell are pushed via queues | 18:46 |
| *** dachary has quit IRC | 18:47 | |
| comstud | API will update its state based off that | 18:47 |
| comstud | but it's always slightly stale | 18:47 |
| comstud | even today in non-cells.. it's slightly stale | 18:47 |
| comstud | or can be | 18:47 |
| maoy | stale meaning? | 18:47 |
| comstud | i mean | 18:47 |
| comstud | well | 18:47 |
| comstud | as soon as you return from an API call, the state is no longer valid | 18:47 |
| comstud | but also... | 18:47 |
| comstud | there's no transactions | 18:47 |
| maoy | vm_state or task_state? | 18:47 |
| comstud | getting the state | 18:47 |
| comstud | and setting the state are not atomic | 18:48 |
| comstud | both | 18:48 |
| *** Ryan_Lane has joined #openstack-dev | 18:48 | |
| *** Ryan_Lane has joined #openstack-dev | 18:48 | |
| maoy | it depends on how you look at this. | 18:48 |
| maoy | say you creating a VM, when the task finishes on the worker, the task is done in the physical world | 18:49 |
| comstud | you're proposal here is a big API change | 18:49 |
| maoy | but not from the db's perspective | 18:49 |
| comstud | tasks finishing and not being marked as finished immediately isn't so much of a problem | 18:50 |
| *** sdake has joined #openstack-dev | 18:50 | |
| maoy | i actually think that only the state machine transition related part is affected | 18:50 |
| maoy | they are a bit broken anyway | 18:50 |
| comstud | anyway... | 18:50 |
| maoy | the other change is how you translate vm_state + task_state to the ec2 state | 18:51 |
| comstud | I think what you have here makes sense | 18:51 |
| comstud | as far as the DB is concerned | 18:52 |
| comstud | mapping them back to what the API returns is something else | 18:52 |
| comstud | like you just said w/ ec2 | 18:52 |
| comstud | hehe | 18:52 |
| maoy | ok. | 18:53 |
| comstud | not sure about 'INITIALIZED', but | 18:53 |
| comstud | this is probably good for discussion | 18:53 |
| maoy | sure | 18:53 |
| comstud | but it sounds like you're talking about now adding a subtask_state too | 18:53 |
| comstud | so that'll make 4! | 18:53 |
| comstud | hehe | 18:53 |
| maoy | state machine should ignore subtask_state | 18:54 |
| maoy | that's for something else | 18:54 |
| comstud | yeah | 18:54 |
| comstud | i'm not disagreeing with it | 18:54 |
| comstud | I think they all have valuable meaning | 18:54 |
| maoy | i'm don't completely understand soft-delete and resize | 18:55 |
| maoy | so the state there might be wrong | 18:55 |
| *** joesavak has joined #openstack-dev | 18:56 | |
| *** joesavak has left #openstack-dev | 18:56 | |
| *** joesavak has quit IRC | 18:57 | |
| *** sdake has quit IRC | 18:57 | |
| maoy | btw, what's the preferred way to present stuff like this, google doc vs etherpad vs wiki vs email? | 18:57 |
| *** sdake has joined #openstack-dev | 18:57 | |
| *** rnirmal_ has joined #openstack-dev | 19:02 | |
| *** rnirmal has quit IRC | 19:03 | |
| *** rnirmal_ is now known as rnirmal | 19:03 | |
| comstud | maoy: nod.. sorry, gotta run for now | 19:03 |
| comstud | maoy: I think a lot of us tend to use etherpad more than anything else.. | 19:04 |
| comstud | that's certainly my preference, anyway | 19:04 |
| comstud | :) | 19:04 |
| comstud | if ther'es brainstorming going on, anyway | 19:04 |
| comstud | wiki for links from blueprints, though | 19:04 |
| comstud | i'll bbs | 19:04 |
| *** fc__ has quit IRC | 19:05 | |
| *** torgomatic has joined #openstack-dev | 19:06 | |
| *** novas0x2a|laptop has quit IRC | 19:06 | |
| *** termie has quit IRC | 19:07 | |
| *** adjohn has quit IRC | 19:07 | |
| sdague | I'm trying to add a brand new unit test (new .py file) into nova, but when I do "./run_tests.sh test_virt_driver_loader" I'm getting a AttributeError: 'module' object has no attribute 'test_virt_driver_loader' | 19:07 |
| sdague | is there some other place I need to add this? | 19:07 |
| *** adjohn has joined #openstack-dev | 19:08 | |
| *** novas0x2a|laptop has joined #openstack-dev | 19:08 | |
| openstackgerrit | Verification of a change to openstack/glance failed: Add min_ram and min_disk to bin/glance help. https://review.openstack.org/7516 | 19:15 |
| *** torgomatic has quit IRC | 19:15 | |
| *** davidha has quit IRC | 19:15 | |
| *** davidha has joined #openstack-dev | 19:16 | |
| *** dolphm has joined #openstack-dev | 19:16 | |
| *** torgomatic has joined #openstack-dev | 19:16 | |
| *** galstrom has joined #openstack-dev | 19:16 | |
| Blah1 | sdague: u shouldn't need to add it anyplace else | 19:18 |
| Blah1 | sdague: what's the name of the file and where'd u put it | 19:18 |
| sdague | nova/tests/test_virt_driver_loader.py | 19:19 |
| russellb | comstud: got a moment for a sanity check? trying to figure out how this exception could be caused by these changes ... http://paste.openstack.org/show/17837/ | 19:19 |
| Blah1 | sdague: that looks good to me | 19:20 |
| Blah1 | can u pastebin it somewhere? | 19:20 |
| *** dachary has joined #openstack-dev | 19:21 | |
| sdague | Blah1: yes, one minute, I just blew away my venv, once it rebuilds I can | 19:21 |
| Blah1 | sdague : this is what i have in my test file 'nova/tests/test_blah.py': | 19:22 |
| Blah1 | from nova import test | 19:22 |
| Blah1 | class BlahTestCase(test.TestCase): | 19:22 |
| Blah1 | def test_blah(self): | 19:22 |
| Blah1 | self.assertTrue(1) | 19:22 |
| Blah1 | and that seemed to work for me. | 19:22 |
| Blah1 | ./run_tests.sh test_blah | 19:23 |
| *** matwood has joined #openstack-dev | 19:26 | |
| *** torgomatic has quit IRC | 19:26 | |
| *** matwood has quit IRC | 19:26 | |
| *** torgomatic has joined #openstack-dev | 19:27 | |
| *** vincentricci has quit IRC | 19:27 | |
| Blah1 | dprince : is there a way to submit smoke tests to run on a draft branch prior to having it submitted for review? | 19:28 |
| *** sandywalsh_ has quit IRC | 19:29 | |
| *** anotherjesse is now known as anotherjesse_zz | 19:29 | |
| *** anotherjesse_zz is now known as anotherjesse | 19:30 | |
| maoy | comstud: thanks. | 19:30 |
| comstud | russellb: gimme a few | 19:30 |
| comstud | russelb: it's a None version in dispatcher | 19:31 |
| russellb | k np .. i'm trying to figure it out before you have to tell me, heh | 19:31 |
| comstud | russellb: I ran into that too | 19:31 |
| comstud | oh | 19:31 |
| comstud | it's the scheduler | 19:31 |
| comstud | because it has a __getattr__ | 19:31 |
| dprince | Blah1: Yes. There is a way but it involves manually logging into SmokeStack, putting in your branches, and then firing the job. | 19:32 |
| russellb | oh, hasattr for the API_VERSIOn | 19:32 |
| comstud | yeah | 19:32 |
| russellb | and then boom | 19:32 |
| russellb | awesome, thanks | 19:32 |
| comstud | i'd like to ditch that thing in the scheduler | 19:32 |
| comstud | :) | 19:32 |
| comstud | just haven't gotten there yet | 19:33 |
| comstud | it tries to be too magic-y | 19:33 |
| Blah1 | dprince: oh. what's the best way to just make sure my branch passes smoketest prior to submit for review - would running the local one in the 'smoketest' dir be the same result as the one you're running? | 19:33 |
| russellb | it has bitten me more than one on this patchset | 19:33 |
| comstud | yup | 19:33 |
| *** dachary has quit IRC | 19:34 | |
| russellb | but to quote the source, "# NOTE(vish) magic is fun!" | 19:34 |
| Blah1 | dprince: i ask because i can't get my local smoketests to get anywhere near passing | 19:34 |
| dprince | Blah1: You could go that route. Yes. But the Smoke tests aren't the easiest thing to get running. | 19:34 |
| vishy | russellb comstud : it is fun! | 19:34 |
| Blah1 | i would've thoguht devstack would get that all up and running but apparently not | 19:34 |
| dprince | Blah1: Most devs like to dev test things on devstack. | 19:34 |
| Blah1 | i am using devstack | 19:35 |
| russellb | vishy: agreed :) | 19:35 |
| Blah1 | it's still pretty challenging. a lot of stuff are broken even when it says it stack.sh installed successfully | 19:35 |
| vishy | Blah1: the smoketsts are fairly old, exercise.sh has a similar set of tests | 19:35 |
| vishy | Blah1: which should run out of the box on devstack. | 19:35 |
| Blah1 | vishy: ah yes, exercise.sh appeared to run through well. good to know | 19:35 |
| vishy | Blah1: you have to create a special image and run the tests specifying that image to get smoketests to work | 19:36 |
| vishy | (Blah1 after getting euca credentials and sourcing them) | 19:36 |
| Blah1 | but now that i have your attention vishy, with devstack i can upload images via glance image-create just fine but using ec2ools such as cloud-publish-image decryption fails. is that a known issue or just my lucky dev env? | 19:36 |
| vishy | Blah1: perhaps we should have an exercise script that runs the smoketest | 19:37 |
| Blah1 | and also i notice the eucarc in devstack is broken i have to unset SERVICE_TOKEN && SERVICE_PASSWORD to get it to source | 19:37 |
| vishy | Blah1: if you ran exercises and they all passed, then bundling works through euca2ools. There is an exercise for it. | 19:37 |
| Blah1 | oh. interesting | 19:37 |
| Blah1 | i'll have to go back and check that. | 19:38 |
| vishy | Blah1: check how the script does it, you need to get the right keys | 19:38 |
| vishy | exercises/bundle.sh | 19:38 |
| Blah1 | ya you're right. how'd i miss that! thx again | 19:38 |
| maoy | vishy: i'd also get some your early input on the nova state machine simplification proposal before i broadcast it to the public if you have time | 19:38 |
| maoy | vishy: https://docs.google.com/document/d/1nlKmYld3xxpTv6Xx0Iky6L46smbEqg7-SWPu_o6VJws/edit?pli=1 | 19:38 |
| vishy | maoy: ok | 19:38 |
| sdague | Blah1: hmmm... yep, your example works fine, ok, something else isn't right, I'll bang on it a bit | 19:39 |
| Blah1 | Blah1: cool | 19:40 |
| vishy | maoy: looks good so far | 19:40 |
| Blah1 | whoops i meant sdague cool | 19:40 |
| vishy | maoy: I think you should consider keeping lock free in the following way | 19:40 |
| vishy | some of these fields should only be set from the top down (i.e. user requests) | 19:40 |
| vishy | other fields set only from the bottom up (i.e. reports from compute) | 19:41 |
| vishy | that way you don't have timing conflicts | 19:41 |
| maoy | vishy: power_state is bottom up, task_state is top down | 19:41 |
| vishy | the state reported by the api on show should be due to some logic that resolves all the fields into one view | 19:42 |
| vishy | but what about vm_state? | 19:42 |
| vishy | I would say vm_state should only be top down as well | 19:42 |
| maoy | vm_state is also top down | 19:42 |
| maoy | vm_state and task_state can only be changed via compute api call | 19:42 |
| vishy | so is vm_state really the last task that completed? | 19:43 |
| vishy | should it be last_state? | 19:43 |
| maoy | no. | 19:43 |
| vishy | maoy: how would you change the current state of the vm top down? | 19:43 |
| maoy | vm_state is the real state. task_state is more like a "lock" to describe the on-going task | 19:44 |
| maoy | you can only change the vm_state by make API calls. | 19:44 |
| vishy | but it doesn't change from the call | 19:44 |
| vishy | i.e. to get to rescued | 19:44 |
| vishy | it has to finish rescuing | 19:44 |
| vishy | right? | 19:44 |
| maoy | it does. it's updated at the end of the rescue task, | 19:45 |
| vishy | so it looks like the completion of the task updates the vm_state | 19:45 |
| *** dolphm has quit IRC | 19:45 | |
| maoy | at the same time when you set task_state to None | 19:45 |
| maoy | exactly | 19:45 |
| vishy | maoy: so who is responsible for tracking that state? | 19:45 |
| vishy | the others are simple because it is api request or compute report | 19:46 |
| vishy | it is the vm_state that is tricky | 19:46 |
| maoy | whoever is running the task is responsible to update the vm_state at the end of the task. | 19:46 |
| maoy | right now it's the compute worker. | 19:46 |
| maoy | it will be a different guy when we remove db from compute | 19:46 |
| maoy | if the task fails or stuck, the vm state won't change | 19:47 |
| maoy | but the task_state wont change either | 19:47 |
| sdague | Blah1: ok, figured it out, I had bad import way up the chain, and it was just an unhelpful error message | 19:47 |
| Blah1 | sdague: nice! | 19:47 |
| vishy | maoy: so i guess here is my problem. What is the point of vm_state? | 19:47 |
| maoy | without vm_state, when there is no task running, task_state is None. you don't know what's the status of the vm should be | 19:48 |
| vishy | but why do you care? | 19:48 |
| vishy | we know the power state | 19:48 |
| vishy | we know there are no tasks running | 19:48 |
| vishy | so why is this arbitrary state field important? | 19:48 |
| maoy | the power state is the physical world - what's going on, not what the customer want it to be | 19:49 |
| vishy | maoy: so I think you are saying that the purpose of vm_state is to have something to display to the customer | 19:49 |
| maoy | they can be different and when they are, different policies can be used to address that | 19:49 |
| mnaser | from my experience having two states will cause confusion, all that the customer cares is the real power_state (but forgive me, walking into a discussion here) | 19:49 |
| *** legkodymov has joined #openstack-dev | 19:50 | |
| vishy | maoy: I'm just trying to clarify, if we are only using it to display something reasonable to the customer, then we shouldn't ever key code decisions off of it | 19:50 |
| openstackgerrit | Verification of a change to openstack/nova failed: Finish quota refactor. https://review.openstack.org/7048 | 19:51 |
| vishy | maoy: if it is more than just for the customer than I think we are conflating too many things into one field. | 19:51 |
| legkodymov | Has anyone tried to do container syncing with keystone? | 19:51 |
| maoy | vishy: yup. i know it's hard to explain. that's why we need this discussion. | 19:52 |
| maoy | vishy: the vm_state + task_state together is what's running in the customer's head. power state is reality. both have values. | 19:53 |
| *** Stackops-Jorge has joined #openstack-dev | 19:53 | |
| maoy | vishy: i'm making up an example: | 19:54 |
| vishy | maoy: does the state machine use vm_state as an input to decide what is allowed? | 19:54 |
| maoy | vishy: yes. | 19:54 |
| maoy | vishy: e.g. you can't call unrescue if the vm_state is active | 19:55 |
| maoy | it must be unrescue | 19:55 |
| vishy | maoy: in that case i would prefer we simplify it to only include that information | 19:55 |
| vishy | maoy: maybe you have removed all the other ones already | 19:55 |
| maoy | vishy: i'm afraid they all have to be there. | 19:56 |
| maoy | vishy: yeah i did removed some | 19:56 |
| vishy | maoy: ok i don't see any superfluous ones right now | 19:56 |
| vishy | maoy: i think we potentially need a separate field/timestamp for destroyed | 19:57 |
| maoy | vishy: i thought we had that in the db.. | 19:57 |
| vishy | maoy: we have deleted for when the db record is deleted | 19:57 |
| vishy | maoy: i'm thinking for when the user requests a terminate | 19:57 |
| maoy | vishy: ok. sure. | 19:58 |
| *** anotherjesse is now known as anotherjesse_zz | 19:58 | |
| vishy | maoy: assuming that it may take a while for such a thing to happen, it seems like we should have special case for deprovisioning. | 19:58 |
| vishy | maoy: otherwise I think everything looks pretty good | 19:58 |
| vishy | maoy: we may have to have special logic to get the user-displayed states to match the spec though | 19:58 |
| maoy | vishy: i deleted SHUTOFF from vm_state. that one is probably also the one you have doubt about to be included | 19:58 |
| vishy | maoy: aye SHUTOFF / STOPPED seem the same | 19:59 |
| *** anotherjesse_zz is now known as anotherjesse | 19:59 | |
| maoy | vishy: yes. mapping (vm_state, task_state) to ec2 state is trivial | 20:00 |
| maoy | i'm not very faimilar with OS api | 20:00 |
| vishy | maoy: yes it may be a little less trivial, but should be doable regardless | 20:00 |
| *** galstrom has quit IRC | 20:00 | |
| maoy | delete/destroy/terminate, whatever the name is, should have a higher priority so that it can preeempt other running tasks | 20:01 |
| maoy | which needs a kill task support from compute workers. | 20:01 |
| maoy | vishy: oh. and i'm not clear about the soft-delete semantics.. you can recover the soft-deleted vms if you change your mind? | 20:02 |
| vishy | s1rp: ^^ can you explain soft-delete ? | 20:02 |
| maoy | vishy: is it based on timeout or what? | 20:03 |
| vishy | honestly I think all deletes should be soft by default :| | 20:03 |
| vishy | maoy: or at least providers should be able to keep the instance data around for a while and have it cleaned up later | 20:03 |
| kbringard | this is a super good idea | 20:03 |
| maoy | vishy: so after a soft-delete, it's only a shutdown but no cleanup? there is a periodic_task somewhere to clean up later? | 20:05 |
| vishy | maoy: i think so | 20:05 |
| s1rp | vishy: sure, soft-delete just shutdown the VM and puts it in a `soft-delete` state so that it's in accessible to the customer. a periodic_task (in nova/compute/manager) periodically hard-deletes any soft-deleted instances that have expired | 20:05 |
| s1rp | there is also a python-novaclient extension that allows you to force delete or undelete soft-deleted instances | 20:06 |
| maoy | s1rp: thanks. gotta update the state machine to reflect that.. | 20:07 |
| maoy | s1rp: so when i call terminate in ec2 api, is it softdelete or hard? doesn't it count against my quota? | 20:07 |
| s1rp | maoy: right now soft-delete is only available via the Openstack API, though I don't think there is a reason why we couldn't include it in ec2 as well | 20:09 |
| maoy | s1rp: i c. what about the quota question? that has nothing to do with state machine though. :) | 20:09 |
| s1rp | maoy: i seem to recall it counting against your quota (which seems like a bug) | 20:10 |
| *** renier has joined #openstack-dev | 20:10 | |
| s1rp | maoy: the quota stuff is being completely reworked so i wonder if thats still the case, if it is, we should file a bug against that | 20:11 |
| maoy | s1rp: got it. | 20:11 |
| *** anotherjesse is now known as anotherjesse_zz | 20:12 | |
| *** anotherjesse_zz is now known as anotherjesse | 20:12 | |
| *** legkodymov has quit IRC | 20:12 | |
| *** jakedahn_zz is now known as jakedahn | 20:13 | |
| maoy | vishy: the resize/migration part is the last puzzle to me.. | 20:15 |
| maoy | vishy: are user allow to manually confirm resize via api if task_state is in RESIZE_VERIFY? | 20:15 |
| comstud | yes | 20:16 |
| comstud | maoy: yes | 20:16 |
| comstud | maoy: or optionally, they can revert it | 20:16 |
| maoy | comstud: but a periodic task running in the background is also trying to confirm that automatically? | 20:17 |
| *** cdub has quit IRC | 20:17 | |
| comstud | yes, it'll auto-confirm it for the user after a period of time | 20:17 |
| comstud | if they do not | 20:17 |
| maoy | is either VM, src/dst, running during resizing? | 20:19 |
| maoy | is live_migration a special case of resizing or separate magic? | 20:20 |
| *** wiliam has quit IRC | 20:21 | |
| maoy | comstud, vishy: ^^ | 20:23 |
| comstud | the src is shut down to resize | 20:23 |
| comstud | dest won't start until the resize is done | 20:23 |
| vishy | maoy: live-migration and block-migration are special magic | 20:23 |
| vishy | (and they only work on kvm right now) | 20:23 |
| comstud | xenapi coming! | 20:24 |
| comstud | supposedly | 20:24 |
| *** adalbas has quit IRC | 20:24 | |
| maoy | comstud: ok. now the state machine is clearer. but how does the user know when to confirm resize? or whether to confirm? | 20:24 |
| comstud | because they're return a state of resize_verify | 20:25 |
| comstud | or something similar | 20:25 |
| comstud | but also... | 20:25 |
| comstud | they had to have kicked off the resize themselves | 20:25 |
| comstud | so they should know they'd need to confirm it | 20:25 |
| maoy | i c. | 20:25 |
| maoy | have to familiarize myself with those don't exist in ec2.. | 20:26 |
| maoy | what should ec2 api interpret a resize_verify state? | 20:27 |
| maoy | or we assume it doesn't happen. since the user had to use OS api to resize in the first place? | 20:27 |
| *** thingee has joined #openstack-dev | 20:28 | |
| comstud | i try to ignore the ec2 api, honestly. | 20:28 |
| comstud | which is probably bad of me | 20:28 |
| maoy | lol | 20:28 |
| comstud | :) | 20:28 |
| comstud | So I cannot answer your question | 20:28 |
| maoy | fair enough | 20:28 |
| comstud | vishy is a lot more familiar with it than I am | 20:30 |
| *** mcclurmc_ has joined #openstack-dev | 20:34 | |
| vishy | maoy: i think we can safely ignore it in ec2 (just display it as active) | 20:38 |
| vishy | (or running I guess it is in ec2land | 20:38 |
| *** Ruetobas has quit IRC | 20:40 | |
| *** rkukura has joined #openstack-dev | 20:42 | |
| *** Stackops-Jorge has quit IRC | 20:43 | |
| *** vincentricci has joined #openstack-dev | 20:43 | |
| *** johnpostlethwait has quit IRC | 20:44 | |
| *** Stackops-Jorge has joined #openstack-dev | 20:45 | |
| *** davidha has quit IRC | 20:45 | |
| *** mcclurmc_ has quit IRC | 20:45 | |
| *** davidha has joined #openstack-dev | 20:46 | |
| *** lorin1 has joined #openstack-dev | 20:46 | |
| maoy | vishy: how is the ERROR state used in vm_state now? is it safe to say any exception in any task would set it to ERROR? | 20:47 |
| *** johnpostlethwait has joined #openstack-dev | 20:48 | |
| vishy | maoy: not necessarily, for example, an error in a live migrate just rollbacks | 20:48 |
| *** johnpostlethwa-1 has joined #openstack-dev | 20:48 | |
| *** Ruetobas has joined #openstack-dev | 20:48 | |
| maoy | vishy: what can a user do if vm_state is in error besides delete? | 20:49 |
| vishy | maoy: right now nothing. Pretty bad user experience | 20:50 |
| maoy | maoy: i'm proposing to remove ERROR state. the vm_state won't change if the task raises an exception | 20:51 |
| *** devananda has quit IRC | 20:51 | |
| maoy | vishy: ^^ | 20:51 |
| maoy | vishy: this is compatible with migration behavior. | 20:51 |
| maoy | vishy: but i wonder how would the user know if something went wrong. not that hard in general. and we always allow delete no matter what. | 20:52 |
| *** johnpostlethwait has quit IRC | 20:52 | |
| *** littleidea has joined #openstack-dev | 20:57 | |
| vishy | maoy: yeah i don't think you can do that without an api change | 20:57 |
| vishy | maoy: I guess we could just show error to the user if there are faults in the instance_faults table | 20:57 |
| *** stuntmachine has quit IRC | 20:59 | |
| *** andrewsben is now known as andrewsben_zz | 20:59 | |
| maoy | vishy: ok. | 20:59 |
| maoy | vishy: didn't know we had that table. hoho. | 21:01 |
| vishy | maoy: the problem is there are times when things have gone wrong but the instance is still usable (like if a resize fails and is rolled back) | 21:01 |
| maoy | vishy: exactly. | 21:02 |
| vishy | maoy: so it sucks to show the state as ERROR in that case | 21:02 |
| maoy | vishy: if there is a rollback for a task call, and the rollback works, vm_state shouldn't be touched | 21:02 |
| *** galstrom has joined #openstack-dev | 21:02 | |
| maoy | vishy: can we silently add it to instance_fault table? | 21:03 |
| *** troytoman is now known as troytoman-away | 21:03 | |
| vishy | maoy: but the spec says instance_faults only show up in ERROR State | 21:03 |
| vishy | :( | 21:03 |
| maoy | vishy: oops. | 21:03 |
| maoy | vishy: the OS API spec? | 21:04 |
| maoy | vishy: guess we already violate that with livemigration rollback? | 21:05 |
| *** lts has quit IRC | 21:05 | |
| maoy | vishy: anyway, probably a good idea to still keep error state for now.. | 21:06 |
| *** Mkenneth1 has quit IRC | 21:09 | |
| *** renier has quit IRC | 21:09 | |
| *** AlanClark has quit IRC | 21:09 | |
| *** markvoelker has quit IRC | 21:10 | |
| russellb | comstud: updated rpc version patches with your input from earlier .. | 21:10 |
| *** AlanClark has joined #openstack-dev | 21:11 | |
| *** danwent has quit IRC | 21:12 | |
| *** alaski has quit IRC | 21:15 | |
| *** rnirmal has quit IRC | 21:17 | |
| *** lorin1 has quit IRC | 21:19 | |
| comstud | russellb: cool, will take a look shortly | 21:20 |
| *** cmagina has joined #openstack-dev | 21:22 | |
| *** cmagina_ has quit IRC | 21:23 | |
| *** flaviamissi has quit IRC | 21:24 | |
| *** cmagina has quit IRC | 21:24 | |
| *** cmagina has joined #openstack-dev | 21:25 | |
| *** flaviamissi has joined #openstack-dev | 21:26 | |
| *** mcclurmc_ has joined #openstack-dev | 21:26 | |
| *** b1rkh0ff has quit IRC | 21:27 | |
| *** b1rkh0ff has joined #openstack-dev | 21:30 | |
| *** flaviamissi_ has joined #openstack-dev | 21:32 | |
| *** flaviamissi has quit IRC | 21:32 | |
| *** flaviamissi_ is now known as flaviamissi | 21:32 | |
| *** galstrom has quit IRC | 21:33 | |
| *** pixelbeat has quit IRC | 21:36 | |
| *** Mkenneth has joined #openstack-dev | 21:37 | |
| *** renier has joined #openstack-dev | 21:40 | |
| *** ywu has joined #openstack-dev | 21:41 | |
| *** maoy has quit IRC | 21:44 | |
| *** ayoung has quit IRC | 21:45 | |
| *** gyee has quit IRC | 21:45 | |
| *** torgomatic has quit IRC | 21:58 | |
| *** Stackops-Jorge has quit IRC | 21:59 | |
| *** mcclurmc_ has quit IRC | 22:00 | |
| *** sdake has quit IRC | 22:04 | |
| *** Gordonz has quit IRC | 22:05 | |
| *** torgomatic has joined #openstack-dev | 22:05 | |
| *** flaviamissi has quit IRC | 22:06 | |
| *** Mkenneth has quit IRC | 22:06 | |
| *** kbringard has quit IRC | 22:08 | |
| zykes- | jk0: ping | 22:08 |
| *** byeager has quit IRC | 22:09 | |
| openstackgerrit | Verification of a change to openstack/glance failed: Implement image visibility in v2 API https://review.openstack.org/7430 | 22:15 |
| *** Ryan_Lane has quit IRC | 22:15 | |
| *** torgomatic has quit IRC | 22:16 | |
| *** torgomatic has joined #openstack-dev | 22:17 | |
| *** mrunge has quit IRC | 22:18 | |
| *** ecarlin has joined #openstack-dev | 22:18 | |
| *** Mkenneth has joined #openstack-dev | 22:20 | |
| *** salgado has quit IRC | 22:23 | |
| *** PotHix has quit IRC | 22:24 | |
| *** byeager has joined #openstack-dev | 22:25 | |
| *** dachary has joined #openstack-dev | 22:26 | |
| *** ecarlin has quit IRC | 22:29 | |
| *** anderstj_ has quit IRC | 22:33 | |
| *** bhuvan_ has joined #openstack-dev | 22:34 | |
| *** byeager has quit IRC | 22:42 | |
| *** danwent has joined #openstack-dev | 22:44 | |
| *** devananda has joined #openstack-dev | 22:47 | |
| *** jkyle has joined #openstack-dev | 22:50 | |
| *** b1rkh0ff has quit IRC | 22:51 | |
| *** devananda has quit IRC | 22:54 | |
| *** devananda has joined #openstack-dev | 22:55 | |
| *** pmezard has quit IRC | 22:55 | |
| *** dtroyer is now known as dtroyer_zzz | 22:57 | |
| *** littleidea has quit IRC | 23:00 | |
| *** edygarcia has quit IRC | 23:03 | |
| *** gporras has joined #openstack-dev | 23:14 | |
| *** littleidea has joined #openstack-dev | 23:15 | |
| gporras | hey guys, I'm trying to run the keystone tests ( ./run_tests.sh ) but why do I need to have keystoneclient and swift ? is there a way to turn this off and only test whatever I need to test for keystone? | 23:15 |
| *** lloydde has quit IRC | 23:20 | |
| *** mnaser has quit IRC | 23:32 | |
| *** jakedahn is now known as jakedahn_zz | 23:32 | |
| *** mnaser has joined #openstack-dev | 23:35 | |
| *** lloydde has joined #openstack-dev | 23:36 | |
| *** roge has quit IRC | 23:37 | |
| *** flaviamissi has joined #openstack-dev | 23:37 | |
| *** byeager has joined #openstack-dev | 23:39 | |
| gporras | not a usefull channel | 23:40 |
| *** mnaser has quit IRC | 23:40 | |
| *** gporras has left #openstack-dev | 23:40 | |
| *** thingee has quit IRC | 23:46 | |
| *** AlanClark has quit IRC | 23:46 | |
| *** ywu has quit IRC | 23:47 | |
| *** lloydde_ has joined #openstack-dev | 23:51 | |
| *** lloydde has quit IRC | 23:53 | |
| *** mnaser has joined #openstack-dev | 23:54 | |
| *** lloydde_ has quit IRC | 23:57 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!