Wednesday, 2017-06-14

« Tuesday, 2017-06-13 Index Thursday, 2017-06-15 »
openstackgerritJude Cross proposed openstack/octavia master: Add filtering and field selection to API  https://review.openstack.org/46927500:02
JudeCAlright there we go, filtering on the db :)00:04
johnsomWahoo00:05
johnsomI will take a look again tomorrow.  I am deep in the auth/policy stuffs right now.00:05
openstackgerritMichael Johnson proposed openstack/octavia master: Add RBAC enforcement to Octavia v2 API  https://review.openstack.org/47287200:15
johnsomOk, no more trusting non-existent environment variables to tell me if it is an admin00:17
*** JudeC has quit IRC00:34
rm_worklol00:37
*** aojea has quit IRC00:54
*** aojea has joined #openstack-lbaas00:57
openstackgerritMichael Johnson proposed openstack/octavia master: Add RBAC enforcement to Octavia v2 API  https://review.openstack.org/47287201:05
johnsomWell, that patch should have the base load balancer API protected by policy.  Still more work to do, but a good stopping place.01:06
*** cpuga_ has quit IRC01:12
rm_workk I'll definitely take a look when you're ready01:13
rm_workthat will be important for us in the next few weeks01:13
*** cpuga has joined #openstack-lbaas01:13
*** ianychoi has joined #openstack-lbaas01:14
*** cpuga has quit IRC01:17
johnsomOk, catch you folks tomorrow01:21
*** aojea has quit IRC01:25
*** aojea has joined #openstack-lbaas01:26
rm_worksame01:30
*** cpuga has joined #openstack-lbaas02:24
*** sanfern has quit IRC02:27
*** sanfern has joined #openstack-lbaas02:28
*** yamamoto_ has joined #openstack-lbaas02:32
*** aojea has quit IRC02:52
*** links has joined #openstack-lbaas03:14
*** gans has joined #openstack-lbaas03:14
*** dayou has quit IRC03:27
*** dayou has joined #openstack-lbaas03:28
openstackgerritMichael Johnson proposed openstack/octavia master: Add RBAC enforcement to Octavia v2 API  https://review.openstack.org/47287203:35
*** amotoki_away is now known as amotoki03:55
*** gans819 has joined #openstack-lbaas04:15
*** gans has quit IRC04:18
*** kobis has joined #openstack-lbaas04:30
*** kobis has quit IRC04:38
*** gcheresh has joined #openstack-lbaas05:06
openstackgerritChangBo Guo(gcb) proposed openstack/neutron-lbaas master: Remove usage of parameter enforce_type  https://review.openstack.org/47406205:25
*** kobis has joined #openstack-lbaas05:45
*** kobis has quit IRC05:49
*** kobis has joined #openstack-lbaas05:50
*** rcernin has joined #openstack-lbaas06:08
*** gans819 has quit IRC06:18
*** sanfern has quit IRC06:28
*** sanfern has joined #openstack-lbaas06:30
*** cpuga has quit IRC07:15
*** tesseract has joined #openstack-lbaas07:16
*** korean101 has joined #openstack-lbaas07:53
korean101hi guys07:54
korean101again me...07:54
korean101i try to octavia via vagrant on single node (Ocata Releases)07:55
korean101all process is OK. but something wrong (node1 | BUILD  | scheduling | NOSTATE)07:55
korean101my reop status (http://paste.openstack.org/show/n927PY5RNbTdXeT2MSRM/)07:57
korean101ERROR messages (No cells are configured, unable to list instances)07:58
korean101but i try this command already (nova-manage cell_v2 discover_hosts --verbose)07:58
korean101GMT+9 timezone...07:58
korean101different timezone...07:58
*** cpuga has joined #openstack-lbaas08:16
*** dayou has quit IRC08:18
*** cpuga has quit IRC08:21
*** ianychoi has quit IRC08:43
*** yamamoto_ has quit IRC08:44
nmagneziajo, ^ maybe you got the same error back in the day? seems like korean101 is using your vargant files your :-)09:00
ajohmmm korean101 I may need to try them again09:00
ajolong time without trying09:00
ajoit seems to be a problem with nova btw09:01
ajoI never had that issue09:01
ajodoes it work with master?09:01
*** yamamoto has joined #openstack-lbaas09:12
*** yamamoto has quit IRC09:17
*** yamamoto has joined #openstack-lbaas09:20
*** yamamoto has quit IRC09:43
*** ianychoi has joined #openstack-lbaas09:49
*** yamamoto has joined #openstack-lbaas09:58
*** yamamoto has quit IRC10:05
*** yamamoto has joined #openstack-lbaas10:05
*** yamamoto has quit IRC10:05
*** yamamoto has joined #openstack-lbaas10:51
*** yamamoto has quit IRC10:55
*** yamamoto has joined #openstack-lbaas11:01
*** yamamoto has quit IRC11:06
*** cpuga has joined #openstack-lbaas11:18
*** cpuga has quit IRC11:22
*** yamamoto has joined #openstack-lbaas11:36
*** yamamoto has quit IRC11:36
*** yamamoto has joined #openstack-lbaas11:38
*** dayou has joined #openstack-lbaas11:57
*** aojea has joined #openstack-lbaas12:05
*** atoth has quit IRC12:06
*** belharar has joined #openstack-lbaas12:06
*** leitan has joined #openstack-lbaas12:09
*** leitan_ has joined #openstack-lbaas12:10
*** leitan has quit IRC12:10
*** belharar has quit IRC12:15
*** chlong has quit IRC12:20
*** belharar has joined #openstack-lbaas12:33
*** belharar has quit IRC12:45
*** links has quit IRC12:46
*** catintheroof has joined #openstack-lbaas12:52
*** belharar has joined #openstack-lbaas12:53
*** belharar has quit IRC13:00
*** openstackgerrit has quit IRC13:18
*** cpuga has joined #openstack-lbaas13:19
*** aojea has quit IRC13:23
*** cpuga has quit IRC13:24
*** sanfern has quit IRC13:32
*** sanfern has joined #openstack-lbaas13:33
*** aojea has joined #openstack-lbaas13:38
*** leitan_ has quit IRC13:40
*** leitan has joined #openstack-lbaas13:52
*** atoth has joined #openstack-lbaas13:53
*** amotoki is now known as amotoki_away13:59
*** fnaval has joined #openstack-lbaas13:59
*** cpuga has joined #openstack-lbaas14:00
*** chlong has joined #openstack-lbaas14:02
*** cpuga has quit IRC14:04
*** cpuga has joined #openstack-lbaas14:05
*** chlong has quit IRC14:06
*** belharar has joined #openstack-lbaas14:08
*** links has joined #openstack-lbaas14:19
*** sanfern has quit IRC14:25
*** sanfern has joined #openstack-lbaas14:26
*** armax has joined #openstack-lbaas14:27
*** kobis has quit IRC14:28
*** krypto has joined #openstack-lbaas14:31
*** sanfern has quit IRC14:33
*** chlong has joined #openstack-lbaas14:34
*** chlong has quit IRC14:35
*** links has quit IRC14:36
*** aojea has quit IRC14:37
*** chlong has joined #openstack-lbaas14:38
*** aojea has joined #openstack-lbaas14:45
*** aojea has quit IRC14:52
*** aojea has joined #openstack-lbaas14:56
*** dayou has quit IRC14:56
*** krypto has quit IRC15:06
*** krypto has joined #openstack-lbaas15:06
*** fnaval has quit IRC15:14
*** reedip_ has joined #openstack-lbaas15:34
*** armax has quit IRC15:40
*** armax has joined #openstack-lbaas15:40
*** armax has quit IRC15:42
*** sanfern has joined #openstack-lbaas15:44
sanfernhi ,15:46
*** blogan has joined #openstack-lbaas15:46
cpugahello15:47
johnsomo/15:49
cpugajohnsom: haven't had feedback lately on the flavor spec and was wondering if more info was needed15:55
johnsomcpuga Yeah, I asked some folks to review it, but I'm not sure they have.  I will bring it up again at the meeting today and make it a review priority for me today as well.15:56
johnsomSorry, keep nagging in the channel though!15:56
cpugajohnsom: thx15:57
*** yamamoto has quit IRC15:59
*** yamamoto has joined #openstack-lbaas16:01
*** fnaval has joined #openstack-lbaas16:05
*** yamamoto has quit IRC16:07
*** kobis has joined #openstack-lbaas16:07
*** cody-somerville has joined #openstack-lbaas16:17
*** csomerville has joined #openstack-lbaas16:18
*** cody-somerville has quit IRC16:22
*** diltram has quit IRC16:34
*** kobis has quit IRC16:39
*** diltram has joined #openstack-lbaas16:40
*** leitan has quit IRC16:43
*** diltram has quit IRC16:44
*** armax has joined #openstack-lbaas16:46
*** belharar has quit IRC16:48
*** rcernin has quit IRC16:49
*** aojea has quit IRC16:50
*** diltram has joined #openstack-lbaas16:51
*** tesseract has quit IRC16:56
*** rstarmer has joined #openstack-lbaas17:01
*** sshank has joined #openstack-lbaas17:02
*** catintheroof has quit IRC17:02
*** catintheroof has joined #openstack-lbaas17:03
*** yamamoto has joined #openstack-lbaas17:03
*** rstarmer has quit IRC17:07
*** yamamoto has quit IRC17:18
*** gcheresh has quit IRC17:18
*** aojea has joined #openstack-lbaas17:20
*** reedip_ has quit IRC17:21
*** rstarmer has joined #openstack-lbaas17:26
*** pcaruana has quit IRC17:29
*** krypto has quit IRC17:32
*** krypto has joined #openstack-lbaas17:32
*** krypto has quit IRC17:32
*** krypto has joined #openstack-lbaas17:32
*** rstarmer has quit IRC17:34
*** rstarmer has joined #openstack-lbaas17:35
*** rstarmer has quit IRC17:37
johnsomUgh, logged in today to find uwsgi at 100% on all of the cores.  placement-api and keystone.17:38
johnsomI really don't want to move octavia to use this just for that goal17:38
johnsomrestarted them both and back to normal, but, ugh17:38
*** KeithMnemonic has joined #openstack-lbaas17:55
*** yamamoto has joined #openstack-lbaas17:57
*** krypto has quit IRC18:07
*** sshank has quit IRC18:32
rm_workjohnsom: honestly i am ok with saying "no"18:48
rm_workwe did the goal once18:48
rm_workthey can suck it18:48
rm_workmaybe I should say things like that in PMs :P18:49
*** csomerville has quit IRC18:50
*** cody-somerville has joined #openstack-lbaas18:51
*** cody-somerville has quit IRC18:51
*** cody-somerville has joined #openstack-lbaas18:51
*** openstackgerrit has joined #openstack-lbaas18:58
openstackgerritAdam Harwell proposed openstack/octavia master: Also remove amphora_health entry during house_keeping amps  https://review.openstack.org/47431818:58
rm_workjohnsom: ^^ minor bug that was leaving stale entries18:59
rm_workI think18:59
rm_workwonder if you were seeing the same thing18:59
rm_workactually, crap i might need to do that in a try/except18:59
openstackgerritAdam Harwell proposed openstack/octavia master: Also remove amphora_health entry during house_keeping amps  https://review.openstack.org/47431819:05
rm_workthere we go19:06
*** leitan has joined #openstack-lbaas19:08
*** rcernin has joined #openstack-lbaas19:10
*** gcheresh has joined #openstack-lbaas19:22
*** aojea has quit IRC19:34
*** leitan has quit IRC19:39
*** JudeC has joined #openstack-lbaas19:45
*** cpuga has quit IRC19:49
*** cpuga has joined #openstack-lbaas19:49
*** m-greene has joined #openstack-lbaas19:51
*** sshank has joined #openstack-lbaas19:57
*** leitan has joined #openstack-lbaas20:00
johnsomOctavia meeting starting soon on #openstack-meeting-alt20:00
*** atoth has quit IRC20:16
*** fnaval has quit IRC20:21
*** fnaval has joined #openstack-lbaas20:27
*** aojea has joined #openstack-lbaas20:35
*** aojea has quit IRC20:40
*** gcheresh has quit IRC20:42
*** JudeC has quit IRC20:56
*** jniesz has joined #openstack-lbaas21:00
jnieszalso for the logging, need to be careful about logging to local disk.  I know that can kill performance21:01
johnsomI think we could use the existing amp cert to go from syslog in amp to the lb-mgmt-net proxy, then they can manage certs from there....  Keep it easier21:02
johnsomjniesz Yeah, the current config is to disk with rotation21:02
jnieszthoughts on using ramdisk21:02
johnsomgrin, spinning rust is cheaper the RAM in most cases....21:03
johnsomThough with the syslog we could offer a pure streaming solution21:03
jnieszyea, I like the pure streaming, but sometimes useful to have a small amount local in case you are debugging network issue21:04
johnsomIf someone is fancy and has the time you could add that to the amp admin API to flip it on and off21:05
rm_workyeah though RAM is very important for actual loadbalancing21:05
jnieszdisk kills perf,21:05
johnsomYeah, if you have a high rate of connections, definitely21:06
jnieszwhen I was stressing the LB, turning on httplog killed the lb21:06
jnieszwhen it was logging to local disk21:06
jnieszstreaming only have a few ms impact21:06
jnieszhad21:06
jnieszxgerman_ I saw the info you posted for active/active.  Wanted to discuss the distributor role a little bit21:08
johnsomI wonder if there are issues with the syslog engine between redhat tribe and debian tribe.  Do they both use the same package?21:08
xgerman_sure21:08
jnieszthe distributor can live on a VM or can simply exist as a function of the L3 network21:09
xgerman_my worry is that we are introducing a complete new vm life ccle and we could re-use amphora21:09
jnieszin L2 there would be distributor VMs21:09
jnieszin L3 we could have distributor VMs for control plane21:09
jnieszexample BGP route injection21:09
jnieszin L3 data plane will be the network21:10
xgerman_ok, but couldn’t we just manage them like amphora?21:10
jnieszif we keep it separate we could provide UDP load balancing without an amphora21:10
jnieszjust inject anycast routes21:11
jnieszfrom the "distributor" VMs21:11
johnsomThat was my issue with the original act/act patches, it's basically an amp, we should get the lifecycle benefits21:11
xgerman_+121:11
jnieszdoes amphora always imply having a lb (haproxy)21:12
jniesz?21:12
xgerman_no21:12
jnieszor can an amphora just be an empty VM running a bgp speaker21:12
johnsomno, haproxy is not running by default in the amp, only after listener is configured21:13
*** aojea has joined #openstack-lbaas21:13
xgerman_yueah, we could have amphora where no haproxy is installed when we get multi-image right21:13
jnieszwhat about use-case where the distributor is in the data plane?21:14
jnieszwould that be an amp (ovs / openflow) -> amp21:15
jniesz-> multiple amps (haproxy)21:15
jnieszI think it makes sense to refer to that as distributor21:17
johnsomYeah, I think terminology wise it makes sense to differentiate them, but technology, if we are using a service VM, we should leverage as much as possible.21:18
johnsomUnder the old patches, it seems like distributor is really just a type of amp21:19
xgerman_I am pretty sure it will be si ilar with the new model21:19
xgerman_regarding data plane are we trying to talk to routers? Then we could make some driver and just blast directly to them?21:20
jnieszfor each anycast ECMP route you inject you need a unique BGP peer21:21
jnieszthat is why the model had the bgp speaker on the amp21:21
jnieszon each amp in the cluster21:21
xgerman_yep, that would argue for an extension of the amp api21:22
jnieszif you advertise multiple next-hop routes from the same peer session BGP will select only 121:22
xgerman_oh, ok21:22
johnsomOk, I am now confused (haven't read the new spec yet).  What would the distributor do in your model?  Yeah, you have to speak from each amp with an active haproxy right?21:23
jnieszwould need a bgp speaker on each amp peering with the network fabric21:23
jnieszthe distributor is part of the fabric itself21:24
jnieszso it wouldn't live on any compute node21:24
jnieszor in any amp21:24
johnsomRight21:24
johnsomAlternate act/act designs may have a distributor service vm21:24
xgerman_oh, ok — that confused us since the last proposal they were running ovs in a servcie vm and calling it  distributor21:25
jnieszyes21:25
xgerman_so basically we don’;t have to talk to the distributor accept through the BGP speaker21:25
xgerman_but need to know of it’s existance so we can configure the speaker right?21:26
jnieszcorrect, we are talking to the distributor (l3 network) via the speaker on each amp21:26
jnieszneed to then map speaker to amp21:26
xgerman_ok, got it21:26
jnieszthat was initial release21:27
jnieszbut in future we could also do cool stuff like udp load balancing21:27
jnieszwith no software LB, and just use bgp speakers in amps21:27
xgerman_ok21:27
xgerman_that’s not precluded with that design21:27
jnieszcorrect21:28
xgerman_as johnsom said we don’t need to activate the haproxy21:28
johnsomI think that UDP setup could get interesting though as you assume that device has visibility to the tenant networks....21:29
johnsomWe need to think more about that and how it can be flexible for different topologies21:29
jnieszthe network itself should be able to route to all the tenant networks21:29
xgerman_you can always put your ampson public or use FIPs21:29
xgerman_jniesz that’snot a given in some installations21:30
jnieszso amp should only need lb-mgmt-net21:30
jnieszor can take xgerman_'s suggestion and make it optional to peer over tenant network21:30
johnsomYeah, jniesz many don't.  They are all virtual neutron networks.21:31
jnieszfor neutron networks (vxlan) would need FIP21:31
johnsomYeah, you don't with octavia today21:31
johnsomOk, I need to do some patch reboots.  Back in a bit21:32
xgerman_gotta run, too — but I think we are on the right track…21:33
jnieszfor L3 active/active to work either FIP or IPs are routeable directly from the network fabric you are advertising the routes to21:33
jnieszok, I have to sign off for the day as well21:33
jnieszwe can discuss more next time21:33
*** jniesz has quit IRC21:36
rm_worki keep seeing "don't have to activate the haproxy on the amp", but.. if we did "typed" amps, they can use different image tags, so you could have an haproxy-amp or a bgp-amp21:46
rm_worknot just "don't activate haproxy"21:46
*** armax has quit IRC22:09
*** JudeC has joined #openstack-lbaas22:10
johnsomTrue22:13
*** armax has joined #openstack-lbaas22:15
*** chlong has quit IRC22:37
*** fnaval has quit IRC22:37
*** sshank has quit IRC22:42
*** aojea has quit IRC22:43
*** cpuga has quit IRC22:48
*** aojea has joined #openstack-lbaas22:54
*** ghs39ghsggf has joined #openstack-lbaas23:02
*** ghs39ghsggf has left #openstack-lbaas23:02
JudeCxgerman_ I hope you don't mind that I am basically stealing all the code from https://review.openstack.org/#/c/459880/4 and putting it on top of my change for filtering/field selection. Can I pop you in as a co-author on the change so you get credit for it?23:04
*** aojea has quit IRC23:04
*** aojea has joined #openstack-lbaas23:08
*** catintheroof has quit IRC23:09
xgerman_Sure, feel free - saves me a rebasing:-)23:11
*** leitan has quit IRC23:12
*** fnaval has joined #openstack-lbaas23:13
JudeC:P23:13
JudeCcool ty23:13
*** csomerville has joined #openstack-lbaas23:14
*** cody-somerville has quit IRC23:14
*** sshank has joined #openstack-lbaas23:20
*** sshank has quit IRC23:20
openstackgerritJude Cross proposed openstack/octavia master: Add status tree to V2 API  https://review.openstack.org/47439023:32
*** JudeC has quit IRC23:36
« Tuesday, 2017-06-13 Index Thursday, 2017-06-15 »

Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!