Thursday, 2014-09-11

*** xgerman has quit IRC00:05
*** vivek-ebay has quit IRC00:10
*** crc32 has quit IRC00:16
*** johnsom has joined #openstack-lbaas00:21
*** sbalukoff has quit IRC00:26
*** jwarendt has quit IRC00:38
*** ptoohill has joined #openstack-lbaas00:54
*** VijayB has quit IRC01:08
*** VijayB has joined #openstack-lbaas01:08
*** VijayB has quit IRC01:10
*** VijayB_ has joined #openstack-lbaas01:10
*** VijayB_ has quit IRC01:13
*** KunalGandhiEbay has quit IRC01:19
*** ptoohill has quit IRC01:34
*** woodster_ has quit IRC01:35
*** mlavalle has quit IRC01:35
*** ptoohill has joined #openstack-lbaas01:50
*** markmcclain has quit IRC02:05
*** amotoki_ has joined #openstack-lbaas02:10
*** amotoki_ is now known as amotoki02:11
*** jamiem has quit IRC02:20
*** sbfox has joined #openstack-lbaas02:26
*** markmcclain has joined #openstack-lbaas02:26
*** vivek-ebay has joined #openstack-lbaas02:29
*** sbfox has quit IRC02:43
*** ajmiller has quit IRC02:50
*** vivek-eb_ has joined #openstack-lbaas02:59
*** vivek-ebay has quit IRC03:01
*** sbfox has joined #openstack-lbaas03:29
*** sbalukoff has joined #openstack-lbaas04:13
*** markmcclain has quit IRC04:34
*** HenryG is now known as HenryG_zzzz04:40
*** ptoohill has quit IRC04:54
*** ptoohill has joined #openstack-lbaas04:54
*** ptoohill has joined #openstack-lbaas04:54
*** KunalGandhiEbay has joined #openstack-lbaas05:19
*** vivek-eb_ has quit IRC05:54
*** KunalGandhiEbay has quit IRC06:13
*** sbfox has quit IRC06:55
*** jschwarz has joined #openstack-lbaas07:45
*** ptoohill has quit IRC07:54
*** ptoohill has joined #openstack-lbaas07:56
*** ptoohill has quit IRC08:29
*** ptoohill has joined #openstack-lbaas08:36
*** ptoohill has quit IRC08:59
*** ptoohill has joined #openstack-lbaas09:01
*** maishsk has joined #openstack-lbaas09:45
maishskenikanorov: HI - you around?09:45
*** jschwarz has quit IRC09:57
*** jamiem has joined #openstack-lbaas11:21
*** enikanorov_ has joined #openstack-lbaas11:54
*** enikanorov__ has quit IRC11:54
*** RaginBajin has joined #openstack-lbaas12:02
*** markmcclain has joined #openstack-lbaas12:09
*** woodster_ has joined #openstack-lbaas12:09
*** woodster_ is now known as woodster12:10
*** ajmiller has joined #openstack-lbaas12:40
*** HenryG_zzzz is now known as HenryG12:55
*** ptoohill has quit IRC12:56
*** ptoohill has joined #openstack-lbaas12:56
sballe_The website for session suggestions isn't open yet... Any idea when that willl happen?
sballe_mestery, ^^^^13:36
mesterysballe_: Not until after RC I believe13:37
sballe_ok thx13:37
*** vivek-ebay has joined #openstack-lbaas13:48
*** balles has joined #openstack-lbaas13:50
*** samuelbercovici has joined #openstack-lbaas13:58
*** markmcclain has quit IRC13:58
*** xgerman has joined #openstack-lbaas13:58
*** maishsk has quit IRC13:58
*** maishsk_ has joined #openstack-lbaas13:58
*** xgerman has quit IRC13:58
*** xgerman has joined #openstack-lbaas13:58
*** xgerman has quit IRC13:58
*** xgerman has joined #openstack-lbaas13:58
*** xgerman has quit IRC13:59
*** xgerman has joined #openstack-lbaas13:59
*** jorgem has joined #openstack-lbaas13:59
*** amotoki has quit IRC14:21
*** fnaval has joined #openstack-lbaas14:22
*** VijayB_ has joined #openstack-lbaas14:26
*** markmcclain has joined #openstack-lbaas14:47
*** markmcclain1 has joined #openstack-lbaas14:49
*** markmcclain1 has quit IRC14:49
*** balles has quit IRC14:49
*** markmcclain1 has joined #openstack-lbaas14:50
*** maishsk_ has quit IRC14:50
*** samuelbercovici has quit IRC14:51
*** markmcclain has quit IRC14:52
*** balles has joined #openstack-lbaas14:54
*** ptoohill-oo has joined #openstack-lbaas15:00
*** ptoohill has quit IRC15:00
*** jorgem has quit IRC15:01
*** jorgem has joined #openstack-lbaas15:02
*** jorgem has quit IRC15:02
*** jorgem has joined #openstack-lbaas15:03
*** balles has quit IRC15:04
*** balles has joined #openstack-lbaas15:06
*** jorgem has quit IRC15:10
*** jorgem has joined #openstack-lbaas15:13
*** markmcclain1 has quit IRC15:20
*** markmcclain has joined #openstack-lbaas15:20
*** balles has quit IRC15:24
*** balles has joined #openstack-lbaas15:26
*** balles has quit IRC15:50
*** balles has joined #openstack-lbaas15:54
*** mlavalle has joined #openstack-lbaas16:03
*** vivek-ebay has quit IRC16:14
*** markmcclain has quit IRC16:19
*** sbalukoff has quit IRC16:28
*** balles has quit IRC16:35
*** vivek-ebay has joined #openstack-lbaas16:38
*** balles has joined #openstack-lbaas16:38
*** markmcclain has joined #openstack-lbaas17:02
*** ptoohill-oo has quit IRC17:04
*** VijayB_ has quit IRC17:08
*** balles has quit IRC17:35
*** balles has joined #openstack-lbaas17:38
*** sbalukoff has joined #openstack-lbaas17:42
*** openstackgerrit has quit IRC17:46
*** VijayB has joined #openstack-lbaas17:48
*** openstackgerrit has joined #openstack-lbaas17:48
*** mikedillion has joined #openstack-lbaas18:27
*** mikedillion has quit IRC18:30
*** ptoohill has joined #openstack-lbaas18:32
*** crc32 has joined #openstack-lbaas19:02
*** sbfox has joined #openstack-lbaas19:10
*** markmcclain has quit IRC19:28
xgermanblogan, yt?19:34
*** mlavalle_ has joined #openstack-lbaas19:37
*** mlavalle has quit IRC19:40
*** rohara has quit IRC19:40
*** rohara has joined #openstack-lbaas19:40
*** VijayB has quit IRC19:45
*** crc32 has quit IRC19:45
*** jkoelker has quit IRC19:46
bloganxgerman huh?19:47
xgermanour lawyers wanted to know if Octavia enforces the CLA - luckily:
xgermantook me a while and some help to figure that out :-)19:48
bloganxgerman: ahh nice19:51
xgermanyeah, in case you get the same question from your lawyers :-)19:51
*** VijayB has joined #openstack-lbaas19:54
*** mestery has quit IRC19:54
*** mestery has joined #openstack-lbaas19:54
*** sbfox1 has joined #openstack-lbaas20:07
*** sbfox has quit IRC20:07
*** markmcclain has joined #openstack-lbaas20:10
*** jkoelker has joined #openstack-lbaas20:12
dougwigxgerman: here20:20
xgermanI figured it out...20:20
blogani try to stay away from lawyers20:23
bloganif i have to talk to a lawyer 9 times out of 10 its not good20:23
dougwigthen you need to be committing better crimes.20:31
*** vivek-ebay has quit IRC20:43
bloganim going to make a trip to idaho and dress up as an elk, then you'll be the one talking to a lawyer20:45
*** markmcclain has quit IRC20:59
*** mestery has quit IRC21:00
*** enikanorov_ has quit IRC21:00
*** enikanorov_ has joined #openstack-lbaas21:00
*** mestery has joined #openstack-lbaas21:00
sbalukoffJust FYI, y'all: I'm out of the office tomorrow and throughout the weekend.21:04
sbalukoffblogan: Heh! I had a brother-in-law who got arrested for shooting a fake deer next to the road. Guy was a complete douche, and my sister was much better off after he divorced his worthless ass.21:06
sbalukoffNot that I'm biased or anything.21:06
bloganyou sound quite biased actually21:06
bloganit sounds like you didn't like him21:06
bloganbut i could be reading that wrong21:06
*** juliancash_ has quit IRC21:13
*** vivek-ebay has joined #openstack-lbaas21:24
*** sbfox1 has quit IRC21:49
*** sbfox has joined #openstack-lbaas21:49
openstackgerritTrevor Vardeman proposed a change to stackforge/octavia: Initial creation of db models, modules, and tests
*** fnaval has quit IRC22:19
*** ptoohill has quit IRC22:25
openstackgerritTrevor Vardeman proposed a change to stackforge/octavia: Initial creation of repository classes and tests
*** rohara has quit IRC22:32
*** ptoohill has joined #openstack-lbaas22:34
*** VijayB has quit IRC22:36
xgermanTrevorV. blogan: Quick question:22:37
xgermanIt looks like you model Amphora 1->*LBs22:37
*** openstackgerrit has quit IRC22:38
*** openstackgerrit_ has joined #openstack-lbaas22:39
*** openstackgerrit_ is now known as openstackgerrit22:40
*** rohara has joined #openstack-lbaas22:41
xgermanI guess I am cutting into foosball time22:42
xgermanand it looks like many-to-many :-(22:49
TrevorVsorry xgerman23:00
TrevorVits many to many my friend23:00
TrevorVFoosball happened after lunch today, I was just using the restroom23:00
xgermancool. I figured that out.23:01
TrevorVYeah, just figured a confirmation wouldn't be a bad thing in this case :D23:01
xgermanI should have paid more attention since I think LB 1->* Amphorae covers most cases23:01
xgermanaka I am lacking the use case for many LB's on one Amphora :-)23:03
TrevorVblogan was in agreement xgerman, but we went with the many to many just in case23:03
xgermanok, I just noticed when I started reviewing your code... anyhow, we can leave it :-)23:03
TrevorVas rm_work suggested, we decided multiple LBs per amphora would be cost efficient (potentially)23:03
xgermanI am more thinking multiple amphora per LB23:04
davidlenwellmultiple tennents on the same lb?23:05
davidlenwellor same tennents pooled on amphora?23:05
xgermansame tennents pooled23:05
davidlenwellokay good23:05
davidlenwellI was going to start flipping tables over23:05
xgermanbut I still think that's a stretch since I would just sue smaller vms to balance23:05
xgermanto each his own though :-)23:06
davidlenwellI think smaller vm's and keeping things more compartmentalized is a good thing23:06
davidlenwellbut thats just my humble opinion23:06
xgermanand you cna overcommit vms to a physical host, too23:06
rm_workmultiple amphora per LB makes sense because that is our scaling mechanism23:06
rm_workanywho, bbl23:06
xgermanrm_work - we are looking itno that, too23:07
xgermanafter all we are now under your old management23:07
xgermanso yeah, in my opinion many<->many is overkill ;-)23:08
TrevorVxgerman, you can definitely comment that in the patch set 1 :D23:10
TrevorVSorry, I meant in the patch set23:10
*** VijayB has joined #openstack-lbaas23:10
xgermanwell, I just +2 blogans chnage establishing that reality the other day23:10
xgermanbut I might :-)23:10
TrevorVRight, but you can always change a review decision, ya know?  :)23:11
TrevorVMake sure to detail its his fault.23:11
xgermansbalukoff always holds me to my +2s as I am infallible23:12
TrevorVsbalukoff, is a frivolous human being.  We accept him for who he is.23:13
sbalukoffxgerman: I'm starting to become less convinced that the multiple LBs per Amphora is a use case we really need to support.23:14
sbalukoffI know we've got those colocation / aplocation requirements that I've been pushing for.23:15
xgermanyep, and on my side we likely will use heavily multiple amphora per load balancer for scaling23:15
sbalukoffBut I think I might be able to push back on that by pointing out that even "dev" instances can simply use smaller amphoras, and that separation of concerns and simplicity is worth it, outweighing the cost-saving argument.23:16
sbalukoffxgerman: Yes, as will we.23:16
sbalukoffWe'll all have to: TLS termination doesn't scale unless you do.23:16
xgermanok, also you can oversubscribe your nova instance to get cost saving there23:16
sbalukoffxgerman: Which is essentially what our customers have been doing by oversubscribing the load balancer.23:17
xgermanyeah, we do it for them :-)23:18
sbalukoffIn any case, one LB per Amphora is how we should write things initially, so we don't have to worry too much about the logic of how to determine when an amphora is "full"23:18
sbalukoff(Which turns out to be a non-trivial problem to solve.)23:18
xgermanand also nova has a way to determine when an instance is full23:19
sbalukoffxgerman: I'm not sure I would trust it. ;)23:19
xgermanwe shouldn't duplicate what they are doing :-)23:19
xgermanyou can always write patches for them :-)23:19
sbalukoffxgerman: We have a very specialized workload, and generic measures tend not to be intelligent enough for that sort of thing.23:19
*** sbfox has quit IRC23:20
sbalukoffWell, maybe.23:20
sbalukoffThe nice part about a specialized workload is we can use additional metrics that might not be available or meaningful to more generic measures (eg. requests / second, number of simultaneous sessions, etc.)23:21
sbalukoffIf we make those available to something like heat, intelligent auto-scale is the next obvious step.23:21
sbalukoffBut... I digress. :)23:21
sbalukoffLet's make our friggen load balancer work first. :D23:22
xgermanyep. On the other hand I don'r know enough about the nova scheduler to have an informed opinion23:22
sbalukoff(For the simple case.)23:22
xgermancool, it seems we are in agreement :-)23:23
sbalukoffxgerman: I know. Strange, isn't it?23:23
sbalukoffGives me the willies.23:23
xgermanfeels weird, too :-)23:24
xgermanso another unrelated question: Are you envisioning the driver to spin up some thread which listens for all the UDP packest from the amphora or would the controller do that and then hand that to the driver23:25
sbalukoffThe former, I'm thinking. Other amphoras might not work the way we make the haproxy amphora work.23:26
sbalukoff(That is to say, they may not use the UDP packets for healthchecks or whatever.)23:26
xgermanyeah, that makes sense - BUT I haye asynchronous drivers23:27
sbalukoffNo way around it.23:27
sbalukoffNot everything can be synchronous. :P23:27
xgermanI know I lost the vote :-)23:27
xgermannow we have to deal with hard problems like do we use eventlet for multithreading, etc.23:28
sbalukoffYeah. :/23:28
sbalukoffI'm hoping that we can make most things synchronous.23:29
sbalukoffAnd that the async stuff is really the relatively rare exception.23:29
xgermanwell, async and databases is a can of worms23:30
xgermanwhich leads to locking, etc.23:30
sbalukoffAlso, I went over our designs with Devananda yesterday, and he pointed out something with how we're communicating that is... annoying. And takes a bit to explain. But it may mean we might have to alter our communication model with the amphoras.23:30
sbalukoffI will be writing a ML thread on this, because... well, it's better to get this figured out before we write a bunch of code for it.23:31
xgermanis it something which is better explained in webEx?23:31
sbalukoffHAHAHA! Actually, I'm not sure it is.23:33
sbalukoffIt'll probably take a diagram.23:33
sbalukoffAnd a lot of text.23:33
*** jorgem has quit IRC23:33
xgermanok, I will wait in anticipation23:34
sbalukoffIt might take me a few days to write up. (We'll see how much time I have for that in Boise. I'm anticipating "not much" so it'll probably be mid next-week before I can.)23:34
sbalukoffBut! It doesn't alter what we need to do at the core in the controller or amphora or anything.23:34
sbalukoffJust how they communicate.23:34
xgermanok, sounds good23:35
xgermanI have to dream up a socket listener, queue, and worker pool to update the DB with stats and health, Fun, fun...23:36
openstackgerritTrevor Vardeman proposed a change to stackforge/octavia: Initial creation of repository classes and tests
*** ajmiller has quit IRC23:58

Generated by 2.14.0 by Marius Gedminas - find it at!