Tuesday, 2018-01-09

« Monday, 2018-01-08 Index Wednesday, 2018-01-10 »
*** openstackgerrit has joined #openstack-lbaas00:22
openstackgerritMerged openstack/octavia master: Change how path munging works for sphinx docs  https://review.openstack.org/52121100:22
*** longstaff has quit IRC00:54
*** longstaff has joined #openstack-lbaas00:54
*** armax has joined #openstack-lbaas01:25
*** longstaff has joined #openstack-lbaas01:35
*** longstaff has quit IRC01:40
*** longstaff has joined #openstack-lbaas01:50
*** longstaff has quit IRC01:53
*** longstaff has joined #openstack-lbaas01:53
*** jniesz has quit IRC01:54
*** longstaff has quit IRC01:58
*** longstaff has joined #openstack-lbaas01:59
*** longstaf_ has joined #openstack-lbaas02:04
*** longstaff has quit IRC02:07
*** annp has joined #openstack-lbaas02:09
openstackgerritOpenStack Proposal Bot proposed openstack/octavia master: Updated from global requirements  https://review.openstack.org/53196902:23
*** harlowja has quit IRC02:32
openstackgerritlei zhang proposed openstack/octavia master: Fix the incorrect git.openstack.org source URL  https://review.openstack.org/53198402:56
openstackgerritEric Lei proposed openstack/neutron-lbaas-dashboard master: Fix the wrong documentation url in README.rst  https://review.openstack.org/53198503:00
openstackgerrithuangshan proposed openstack/octavia master: Check if it is used when creating a load balancer using vip_port_id  https://review.openstack.org/52506903:10
*** longstaf_ has quit IRC03:30
*** longstaff has joined #openstack-lbaas03:30
*** longstaff has quit IRC03:37
*** longstaff has joined #openstack-lbaas03:39
*** yamamoto has joined #openstack-lbaas03:40
*** longstaff has quit IRC03:41
*** longstaff has joined #openstack-lbaas03:42
*** longstaff has quit IRC03:44
*** gans has joined #openstack-lbaas04:12
*** sanfern has joined #openstack-lbaas04:15
sanfernHi johnsom04:16
openstackgerritSanthosh Fernandes proposed openstack/octavia master: L3 ACTIVE-ACTIVE Data model impact  https://review.openstack.org/52472204:18
*** links has joined #openstack-lbaas05:07
*** gans has quit IRC05:35
*** gans has joined #openstack-lbaas05:38
*** gans has quit IRC05:40
*** gans has joined #openstack-lbaas05:42
openstackgerritMerged openstack/octavia master: Updated from global requirements  https://review.openstack.org/53196905:50
*** Rav has joined #openstack-lbaas06:17
*** links has quit IRC06:35
*** rcernin has quit IRC06:38
*** links has joined #openstack-lbaas06:52
*** gans has quit IRC06:53
*** pcaruana has joined #openstack-lbaas06:56
*** links has quit IRC07:03
*** numans has quit IRC07:11
*** numans has joined #openstack-lbaas07:13
*** links has joined #openstack-lbaas07:14
*** links has quit IRC07:27
*** links has joined #openstack-lbaas07:45
*** slaweq has joined #openstack-lbaas07:47
*** PagliaccisCloud has quit IRC07:49
*** slaweq has quit IRC07:52
*** PagliaccisCloud has joined #openstack-lbaas07:52
*** AlexeyAbashkin has joined #openstack-lbaas07:56
*** slaweq has joined #openstack-lbaas08:00
*** b_bezak has joined #openstack-lbaas08:01
*** armax has quit IRC08:08
*** slaweq has quit IRC08:08
*** tesseract has joined #openstack-lbaas08:20
*** Alex_Staf has joined #openstack-lbaas08:25
openstackgerritHengqing Hu proposed openstack/octavia master: ignore api-ref/build directory  https://review.openstack.org/52238508:33
*** yamamoto_ has joined #openstack-lbaas08:36
*** annp has quit IRC08:38
*** annp has joined #openstack-lbaas08:39
*** yamamoto has quit IRC08:39
openstackgerritlei zhang proposed openstack/neutron-lbaas master: Fix the incorrect git.openstack.org source URL  https://review.openstack.org/53207608:41
*** gans has joined #openstack-lbaas08:53
*** aojea has joined #openstack-lbaas08:59
openstackgerritcaishan proposed openstack/octavia-dashboard master: Fix the incorrect git.openstack.org source URL  https://review.openstack.org/53209509:03
*** gans has quit IRC09:22
*** gans has joined #openstack-lbaas09:25
*** aojea has quit IRC09:42
*** yamamoto_ has quit IRC09:52
*** salmankhan has joined #openstack-lbaas09:56
*** aojea has joined #openstack-lbaas09:56
*** gans has quit IRC10:00
*** sanfern has quit IRC10:02
*** sanfern has joined #openstack-lbaas10:12
*** gans has joined #openstack-lbaas10:12
*** yamamoto has joined #openstack-lbaas10:16
openstackgerritSanthosh Fernandes proposed openstack/octavia master: [WIP] ACTIVE-ACTIVE create distributor flow  https://review.openstack.org/52778410:24
nmagnezio/10:39
*** sanfern has quit IRC10:55
*** b_bezak has quit IRC10:56
*** AlexeyAbashkin has quit IRC11:02
*** AlexeyAbashkin has joined #openstack-lbaas11:02
*** b_bezak has joined #openstack-lbaas11:03
*** gans has quit IRC11:12
*** aojea has quit IRC11:40
*** dayou has quit IRC11:44
*** dayou has joined #openstack-lbaas11:46
*** b_bezak has quit IRC12:00
*** yamamoto has quit IRC12:06
*** yamamoto has joined #openstack-lbaas12:06
*** annp has quit IRC12:08
*** aojea has joined #openstack-lbaas12:16
*** aojea_ has joined #openstack-lbaas12:22
*** aojea has quit IRC12:24
*** aojea has joined #openstack-lbaas12:27
*** aojea_ has quit IRC12:30
*** aojea_ has joined #openstack-lbaas12:32
*** aojea has quit IRC12:35
*** aojea has joined #openstack-lbaas12:37
*** aojea_ has quit IRC12:40
*** sanfern has joined #openstack-lbaas12:41
*** aojea_ has joined #openstack-lbaas12:42
*** b_bezak has joined #openstack-lbaas12:45
*** aojea has quit IRC12:45
*** aojea has joined #openstack-lbaas12:48
*** aojea_ has quit IRC12:51
*** aojea has quit IRC12:55
*** atoth has joined #openstack-lbaas13:06
*** Rav has quit IRC13:32
*** AlexeyAbashkin has quit IRC13:52
*** b_bezak has quit IRC13:53
*** AlexeyAbashkin has joined #openstack-lbaas14:00
*** aojea has joined #openstack-lbaas14:04
*** b_bezak has joined #openstack-lbaas14:18
*** gans has joined #openstack-lbaas14:21
*** yamamoto_ has joined #openstack-lbaas14:37
*** yamamoto has quit IRC14:40
*** gans has quit IRC14:58
*** longstaff has joined #openstack-lbaas15:12
xgerman_o/15:18
*** armax has joined #openstack-lbaas15:21
openstackgerritJeffrey Longstaff proposed openstack/octavia master: Provider driver spec  https://review.openstack.org/50995715:24
*** sanfern has quit IRC15:24
dayouo/15:25
*** links has quit IRC15:36
*** salmankhan has quit IRC15:38
*** salmankhan has joined #openstack-lbaas15:40
*** slaweq has joined #openstack-lbaas15:56
*** aojea has quit IRC15:56
*** aojea has joined #openstack-lbaas16:07
*** longstaff has quit IRC16:31
cgoncalvesbcafarel, nmagnezi: do you know why https://review.openstack.org/#/c/526380/ hasn't got merged yet?16:40
*** yamamoto_ has quit IRC16:40
xgerman_is it on top of another change?16:42
bcafarelcgoncalves: its parent is https://review.openstack.org/#/c/522626/ so waiting for that one to go in16:42
bcafarelor what xgerman_ said :)16:42
cgoncalvesbcafarel: ah! should have added Depends-On to hint scatterbrains like me ;-)16:45
*** longstaff has joined #openstack-lbaas16:57
*** longstaff has quit IRC16:59
*** aojea has quit IRC17:00
*** longstaff has joined #openstack-lbaas17:00
*** aojea has joined #openstack-lbaas17:00
*** aojea has quit IRC17:00
*** longstaff has quit IRC17:01
*** longstaff has joined #openstack-lbaas17:01
*** slaweq has quit IRC17:02
*** AlexeyAbashkin has quit IRC17:02
openstackgerritMerged openstack/octavia master: Check if it is used when creating a load balancer using vip_port_id  https://review.openstack.org/52506917:03
bcafarelcgoncalves: nah it's better to force people to scratch their head a bit :)17:03
*** Alex_Staf has quit IRC17:04
*** armax has quit IRC17:05
*** sanfern has joined #openstack-lbaas17:28
sanferno/17:28
*** yamamoto has joined #openstack-lbaas17:41
*** drwahl has joined #openstack-lbaas17:42
drwahldoes octavia support a deployment type where multiple public IPs are on a single lbass VM?17:43
johnsomMultiple public IPs for the VIP?17:44
xgerman_you can assign multiplpe FloatingIps17:44
johnsomYeah, if it's the VIP you are talking about, you would have to use floating IPs, the answer would be no.17:44
drwahlessentially, i'm wanting to have multiple public IPs on the front end and multiple VMs on the backend, but only a single lbass VM17:45
drwahlso not a 1-to-many mapping, but a 1-to-1 mapping, but all on a single lbass VM17:45
*** rods_ has joined #openstack-lbaas17:45
drwahlso foo.example.com goes to 1.1.1.1, bar.example.com goes to 1.1.1.2, but all of that is managed in a single instance17:46
johnsomAh, so multiple load balancers per VM.  Currently, no, it's one VM per load balancer.17:46
johnsomYou can stack hostnames on one LB/VM, but they would all need to point to the same IP.17:46
drwahlbased on some research, it looks like the F5 driver allows for multiple lbs per VM. so is it safe to assume that it's not so much a "design" problem, but simply a matter of the way the haproxy driver is implemented?17:47
drwahlwell, to be fair, the f5 driver isn't a "VM" :)17:47
johnsomRight, F5 is not using VMs, in fact, only the Octavia driver is currently using VMs.17:48
johnsomThe API doesn't have any concept of the backend implementation (i.e. VMs, etc.).17:48
johnsomThose are all driver implementation issues.17:49
*** yamamoto has quit IRC17:49
drwahlso maybe another way to ask this is: aside from F5, do you know of any drivers that would provide the feature i'm looking for?17:49
drwahlwould love to do software loadbalancing instead of hardware (hardware is $$$), but i'm having a hard time finding anything that's already published that does that17:50
drwahlthinking we may have to write something ourselves, which isn't ideal...17:50
johnsomNo, there really isn't a driver that stacks load balancers on one VM today.17:50
drwahlis there a reason that has been avoided? am i missing a concept that might bite me in the ass if i were to try something like that?17:51
johnsomWe have talked about allowing that in the octavia driver as an alternate driver, but we don't have use cases for it.17:51
drwahlour use case is: we're using k8s to host many (2000 or so) disparate wordpress instances (one for each customer). k8s supports the concept of LB, and even has an openstack driver for it, but having 2k loadbalancers on top of the rest of the WP reqs is quite a bit of overhead17:53
johnsomWell, mostly performance, fault isolation, lack of use cases.  If you have multiple VIPs (that simple floating IPs don't meet), usually you want those to be performant and isolated from faults in other LBs.  The overhead of the extra VMs is so small it usually doesn't matter.17:53
johnsomSo, do these WP instances really need separate VIP IPs?  Usually these are deployed with multiple hostnames and one VIP IP, saving public IPs.17:54
drwahlthose are good questions :)17:55
drwahlones we're still struggling to find a definitive answer to17:55
johnsomLoad balancers, no matter F5 or octavia, should be thought of as a performance/isolation unit.  There is overhead for each, and if you don't need the performance, just stack them on one LB and use the hostnames/certs to split them out.17:56
*** tesseract has quit IRC17:57
drwahlthanks for the input johnsom. i really appreciate it!17:58
johnsomNP17:58
*** longstaff has quit IRC18:04
*** longstaff has joined #openstack-lbaas18:06
*** longstaff has quit IRC18:08
*** aojea has joined #openstack-lbaas18:11
*** longstaff has joined #openstack-lbaas18:12
*** aojea_ has joined #openstack-lbaas18:15
*** aojea has quit IRC18:17
*** aojea has joined #openstack-lbaas18:19
openstackgerritSanthosh Fernandes proposed openstack/octavia master: L3 ACTIVE-ACTIVE Data model impact  https://review.openstack.org/52472218:21
*** aojea_ has quit IRC18:22
*** aojea_ has joined #openstack-lbaas18:25
*** AlexeyAbashkin has joined #openstack-lbaas18:26
*** aojea has quit IRC18:28
rm_workdrwahl: yeah, shouldn't you be able to use SNI and l7 rules and different pools, to do exactly that?18:28
rm_workdon't need multiple LBs technically18:28
rm_workif a request comes in from a.mysite.com -> direct to pool A (1.1.1.1)18:29
rm_workif a request comes in from b.mysite.com -> direct to pool B (1.1.1.2)18:29
*** aojea has joined #openstack-lbaas18:29
rm_workusing multiple frontend IPs is not relevant for this18:29
rm_workI believe this should be possible with Octavia as-is18:30
*** AlexeyAbashkin has quit IRC18:30
rm_workyou said you're talking about the k8s proxy, right? that is exactly how we do things internally (not with Octavia technically, but with haproxy)18:30
rm_workoctavia could just as well make the exact same haproxy config though18:31
rm_workSNI to different backends based on hostname18:31
*** aojea_ has quit IRC18:32
rm_workor if it's just HTTP, I believe you could use L7 rules to do it without SNI18:32
rm_workthough I assume it's going to be HTTPS18:32
johnsomHTTP supports the hostname method naitve too18:32
rm_workyeah, that's the L7 rules bit, right?18:33
johnsomYou don't need to do anything on the LB side if the members can handle those hosts18:33
rm_worki think he means there's different members per hostname tho18:33
rm_workbecause if this is the k8s proxy, it's acting as a gateway for the different pods interally18:34
rm_workso it does need to do the splitting18:34
drwahlrm_work: ya, that's certainly a possibility18:34
*** aojea_ has joined #openstack-lbaas18:35
drwahlthere is an argument that providing seperate public IPs per customer has value though18:35
drwahlso it's a matter of managing that ask18:35
rm_workso, we dealt with that argument at RAX18:35
rm_workand deemed it to be complete nonsense18:35
drwahlya, dreamhost hasn't gotten there quite yet :)18:35
drwahlmaybe soon18:35
rm_workask someone to actually tell you one real value-add18:36
rm_workand see if they can :P18:36
*** armax has joined #openstack-lbaas18:36
rm_workin our experience, everything people could come up with was actually doable without separate IPs, they just didn't understand how until we explained it :/18:36
rm_workthat said, if you *want* different public IPs, there's no reason you'd want to stack LBs on one VM18:37
rm_workyou'd want as many tiny LBs as you can get18:37
rm_workwhich is exactly the purpose of Octavia really :P18:37
rm_work"being forced to only use one IP" is actually the limiting factor for scale18:38
*** aojea has quit IRC18:38
rm_workthe second you remove that limitation, things get a whole lot better18:38
drwahlthe overhead cost of running that many LBs is concerning for us (which is fair, i think). but i honestly think the counter point is: then you don't need different public IPs18:38
rm_workso, these LBs should be able to run in 2-4G of RAM on a cloud you hopefully already have18:39
rm_workso the math is pretty simple18:39
drwahlya, we already have a cloud18:39
rm_workfind out how many LBs you could reasonably fit on an F518:39
rm_workmultiply that by 4G to see how much RAM you'd need18:39
drwahland ya, 2-4x2000 ends up being a large number :)18:39
drwahlin terms of money18:39
rm_workyeah18:39
rm_workwell, how much is an F5? :P18:40
drwahlalso expensive ;)18:40
rm_workIME "a large number" lol18:40
*** aojea has joined #openstack-lbaas18:40
*** aojea has quit IRC18:40
rm_workso yeah, 8000GB of RAM costs more than an F5? get an F5, is probably the answer IMO18:40
rm_workthough you really can get away with 4000GB18:40
rm_workwe run in 1G*2 pairs18:41
drwahlthere is a non-zero cost associated with each public IP as well18:41
rm_workthough if you support TLS-Termination probably it's safer to go up to 2G18:41
drwahlwhich is another argument to try to widdle those down18:41
johnsomDon't forget the RAM gets de-duped so really saves a lot of RAM since the AMPs are all the same18:41
rm_workyeah18:41
rm_workwait, does it? lol18:41
rm_worki don't see how that's possible18:42
drwahlKSM18:42
drwahldoes de-dupe (some) RAM18:42
*** aojea has joined #openstack-lbaas18:42
rm_workreally?18:42
rm_worki mean they're KINDA the same, but18:42
rm_workthat's some black magic :P18:42
rm_workneed to read up on that18:42
drwahlit's certainly a dark art18:42
johnsomYeah, vmware does a pretty good job in ESX18:42
*** aojea_ has quit IRC18:43
*** longstaff has quit IRC18:43
johnsomSadly workstation does not have that feature....18:43
rm_workhmm18:43
drwahlKVM here, so it's built-in to linux18:43
rm_work"Allows contamination of cryptographic resources in other virtualized guests via the memory row hammer attack" is a little scary18:44
*** slaweq has joined #openstack-lbaas18:44
rm_worki'd be curious to see IRL how that works out for amphorae18:44
rm_worklike, if we had a HV with only amps on it, how much memory would end up being shared18:45
drwahlit's a bit of a lottery18:45
drwahldepending on page sizes and such18:45
*** aojea_ has joined #openstack-lbaas18:45
*** slaweq has quit IRC18:46
*** irenab has quit IRC18:48
*** oanson has quit IRC18:48
*** aojea has quit IRC18:48
johnsomrm_work the two spectre exploits across processes is a bit scarier to me18:49
rm_worklol18:49
rm_workyes fair18:49
*** aojea has joined #openstack-lbaas18:50
johnsomSide comment, our gates still run with 1GB RAM vms, so really, depending on the load, it can be kinda small18:52
*** aojea_ has quit IRC18:53
rm_workjohnsom: replied on https://review.openstack.org/#/c/525353/18:53
rm_workand I think we can ignore the -1 on https://review.openstack.org/#/c/525778/18:54
johnsomYeah, ok18:55
*** aojea_ has joined #openstack-lbaas18:55
sanfernjohnsom, do we support multiple VIP's on same amphorea ?18:55
johnsomsanfern No, not currently18:55
rm_worksanfern: one VIP, but you can point tons of FLIPs if you want :P18:56
rm_workthere's no good reason to support multiple VIPs IMO18:56
sanfernThanks18:56
johnsomYeah, it would be the odd case.  Likely a driver alternative.  I mean maybe if someone wants to have two massive boxes and load up, but...  Issues18:57
rm_workif you have two massive boxes...18:57
rm_workmake them be hypervisors <_<18:57
johnsomYeah, or you might as well buy an appliance18:57
rm_workand make a LBaaS AZ18:57
*** aojea has quit IRC18:58
*** aojea has joined #openstack-lbaas19:01
*** aojea_ has quit IRC19:04
*** b_bezak has quit IRC19:05
*** aojea_ has joined #openstack-lbaas19:06
*** aojea has quit IRC19:09
*** aojea has joined #openstack-lbaas19:12
*** aojea_ has quit IRC19:15
*** aojea_ has joined #openstack-lbaas19:16
*** aojea has quit IRC19:19
*** aojea has joined #openstack-lbaas19:22
*** salmankhan has quit IRC19:22
*** sanfern has quit IRC19:22
*** sanfern has joined #openstack-lbaas19:22
*** sanfern has quit IRC19:23
*** aojea_ has quit IRC19:25
*** aojea has quit IRC19:25
*** harlowja has joined #openstack-lbaas19:33
*** harlowja_ has joined #openstack-lbaas19:41
*** harlowja has quit IRC19:43
*** dougwig has joined #openstack-lbaas19:54
*** harlowja_ has quit IRC19:55
*** harlowja has joined #openstack-lbaas19:55
*** b_bezak has joined #openstack-lbaas20:01
rm_workjohnsom: wait, i thought we fixed the /proc/net/psched error issue on the docs build20:05
rm_worki just got it again20:05
*** rods_ has left #openstack-lbaas20:05
johnsomUmmm, rings a bell, but not sure on a fix.  Can you paste the error?20:06
*** b_bezak has quit IRC20:06
openstackgerritAdam Harwell proposed openstack/octavia master: Add api-ref for amphora failover  https://review.openstack.org/53230020:08
rm_workjohnsom: ^^ there's that BTW20:09
rm_workhttp://paste.openstack.org/20:10
rm_workand no, scrolling up, nothing else mentions this20:10
rm_workit's kinda out of the blue20:10
rm_worki SWEAR we fixed it tho20:10
rm_workoh hold on20:10
rm_workmaybe this is too old20:10
*** yamamoto has joined #openstack-lbaas20:11
rm_workjust realized how many commits back that patch was20:11
johnsomThat past link is bad20:11
*** slaweq has joined #openstack-lbaas20:17
rm_workerrr20:22
rm_worklol wtf20:22
rm_workhttp://paste.openstack.org/show/641534/20:23
rm_workthere you go, heh20:23
rm_worknot sure how i only got the base URL :P20:23
rm_workbbl showering20:23
rm_workand heading to lunch20:23
rm_work(though I guess the channel didn't need to know that first part)20:23
johnsomYeah, having a sandwich myself20:23
*** yamamoto has quit IRC20:48
*** yamamoto has joined #openstack-lbaas20:49
*** armax has quit IRC20:53
*** reedip has quit IRC20:53
*** yamamoto has quit IRC20:54
*** salmankhan has joined #openstack-lbaas20:57
openstackgerritMerged openstack/octavia master: Amphora API Failover call  https://review.openstack.org/52577821:00
*** salmankhan has quit IRC21:02
*** reedip has joined #openstack-lbaas21:05
*** armax has joined #openstack-lbaas21:34
*** pcaruana has quit IRC21:38
*** slaweq has quit IRC21:39
*** threestrands has joined #openstack-lbaas21:44
*** threestrands has quit IRC21:44
*** threestrands has joined #openstack-lbaas21:44
*** yamamoto has joined #openstack-lbaas21:52
*** aojea has joined #openstack-lbaas21:54
*** yamamoto has quit IRC21:59
*** aojea_ has joined #openstack-lbaas21:59
*** AlexeyAbashkin has joined #openstack-lbaas22:00
*** aojea_ has quit IRC22:02
*** aojea has quit IRC22:02
*** AlexeyAbashkin has quit IRC22:05
*** rcernin has joined #openstack-lbaas22:06
*** AlexeyAbashkin has joined #openstack-lbaas23:22
*** AlexeyAbashkin has quit IRC23:27
*** atoth has quit IRC23:57
« Monday, 2018-01-08 Index Wednesday, 2018-01-10 »

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