Wednesday, 2016-02-03

openstackgerritAdam Gandelman proposed openstack/astara: WIP Add clustering capabilities to instance manager
openstackgerritAdam Gandelman proposed openstack/astara: Rename self.driver to self.resource where applicable
*** shashank_hegde has quit IRC01:32
*** phil_h has quit IRC02:08
*** reedip has joined #openstack-astara03:04
*** shashank_hegde has joined #openstack-astara03:42
*** xiayu1 has joined #openstack-astara06:12
*** xiayu has quit IRC06:13
*** xiayu has joined #openstack-astara06:20
*** xiayu1 has quit IRC06:20
*** xiayu1 has joined #openstack-astara06:26
*** xiayu has quit IRC06:27
*** shashank_hegde has quit IRC08:25
*** xiayu has joined #openstack-astara08:54
*** xiayu has left #openstack-astara08:54
*** xiayu1 has quit IRC08:54
*** phil_h has joined #openstack-astara12:25
*** phil_h has quit IRC12:26
*** phil_h has joined #openstack-astara12:27
*** openstackgerrit has quit IRC13:02
*** openstackgerrit has joined #openstack-astara13:03
*** shashank_hegde has joined #openstack-astara16:23
rodsadam_g you around?16:41
adam_grods, hey16:41
rodsadam_g hi16:41
rodsI'm talking with drwahl to see how we can do some worker monitoring with the rug16:42
rodsadam_g do you have a few minutes to join us?16:42
adam_grods, g+ is probably not doable right now. im baby watching16:43
adam_grods, what do you mean by worker monitoring?16:43
drwahlis that at all like bird watching?16:43
rodsadam_g np, we can chat about it later16:43
adam_gdrwahl, ya, im hiding in the bushes right now16:43
drwahladam_g: we're conceptualizing a monitoring API for the rug16:43
adam_grods, ive actually been thinking of ways to get some visibility into whats happening in the workers, from the outside16:43
drwahlthat reports some metrics about things like how many messages a thread has processes since <some interval>16:44
adam_gand current queue length and such16:44
adam_gessentially the output you get in logs from 'astara-ctl workers debug'16:44
adam_gbut available via the API horizon uses16:44
rodsadam_g cool, we where thinking about something like that too16:45
drwahlat $previous_employer, we had a .jsp page that every process generated and our monitoring systems could poll for data. something similar with an API would be fantastic16:45
adam_gpumping something similar to that through the API and dumping as a json blob of stats would be an easy first step i think16:46
rodsyeah, that would be a good start16:46
drwahlthat would get us at least 80% of what we're looking for16:46
adam_gone thing to consider is that there'd be multiple endpoints to poll if running multiple orchestrators16:47
drwahlsome other interesting metrics (from my perspective) would be total memory allocated, total routers being managed, and average time a message is in the queue16:48
adam_gdrwahl, we already report on the number of state machines managed, the rest of that would need to be implemented somehow16:50
*** phil_h has quit IRC16:50
rodsadam_g do you have a bp for the things we just taked about16:52
rodsdrwahl and I have time to work on it16:52
rodsand would be nice to have something working for the mitaka cycle16:52
adam_grods, i havent created anything, just an idea ive had kicking around to make some progress on this integratin test suite16:53
rodsok, so maybe later if you have time we can chat quickly and drwahl and I can start writing some code16:54
*** phil_hop has quit IRC16:59
*** phil_h has joined #openstack-astara17:04
*** phil_hop has joined #openstack-astara17:17
openstackgerritAdam Gandelman proposed openstack/astara: Adds a functional test for router interfaces
openstackgerritAdam Gandelman proposed openstack/astara: Add astara-ctl + API functional tests
openstackgerritAdam Gandelman proposed openstack/astara: fail func tests early (do not merge)
openstackgerritAdam Gandelman proposed openstack/astara: Adds a functional test for router interfaces
openstackgerritAdam Gandelman proposed openstack/astara: Add astara-ctl + API functional tests
openstackgerritAdam Gandelman proposed openstack/astara: fail func tests early (do not merge)
*** shashank_hegde has quit IRC17:50
*** shashank_hegde has joined #openstack-astara18:14
openstackgerritShilla Saebi proposed openstack/astara: minor changes to orchestrator astara file
openstackgerritmark mcclain proposed openstack/astara: clean up Nova policy changes
*** puranamr has joined #openstack-astara19:58
rodsadam_g are you available for that chat?20:05
adam_grods, sure20:06
*** puranamr has quit IRC20:06
rodsgreat, thx20:06
rodsadam_g you should be able to join us here
phil_hadam_g, do you have a minute?20:38
adam_gphil_h, hi20:44
phil_hhello, getting some problems with setting up load balancers20:44
phil_hhave one related to lb and one astara problem20:45
phil_hthe LB problem - astara launched a LB vm but didn't configure it20:45
phil_hno config for nginx is there20:45
phil_hastara doesn't show any errors though20:46
adam_gphil_h, did you only create the LB and not the members/pool/listeners? it will wait till all of the config exists on the logical resource in neutron before actually configuring the server in the appliance20:46
phil_hI though I set all of that but let me double check20:47
phil_hsomething happened to the members, one I got the added it is working.20:52
phil_hthanks I have another problem but give me minute20:53
adam_gcool, good to hear20:54
phil_hare you testing using IPv4 or IPv^ addresses?20:58
phil_hIf I have astara set up neutron creates a IPv6 subnet whenever I create a network20:59
phil_hwhich is different but I have a problem with IPV4 stuff and getting the l2pop set up on the proper interface21:01
markmcclainright.. there's a setting in astara-neutron to auto create resources include v6 subnets21:07
markmcclainphil_h: ^ we've pushed a change to make this optional with the goal to turn it off before mitaka closes21:07
markmcclainphil_h: the auto resource creation generates too many surprises21:08
phil_hopLet me look and see if that fixes the problems I have seen21:14
elopezwill need to back port the auto create resources to liberty21:19
adam_gmarkmcclain, why do we configur() in check_boot() ?
markmcclainbecause if we can't push an initial config the boot likely failed22:24
adam_gmarkmcclain, why test that with a config push and not is_alive() ?22:25
markmcclainwell a create event only boots an image22:25
markmcclainwe'd have to split it into 222:25
markmcclaina create+update22:25
adam_gwas always wondering why i'd see 2 config pushes on a new instance22:26
adam_gjust stumbled on that22:26
markmcclainthe 2nd push might be a mistake22:27
adam_gfor this clustering stuff, checkboot is going to really only have to check the boot/liveliness, and let ConfigureInstance() actually do the confugiring22:27
markmcclainyeah really all depends on how we want to determine a successful boot22:27
markmcclainthe HV might say the instance is active but if we can't push a config might be a sign that it's borked22:28
adam_gso configure() gets called once from instance_manager's check_boot, and also during ConfigureInstnace()'s execute()22:28
* markmcclain isn't convicted on this subject22:28
adam_gto me, is_alive() should be a sufficien t test of the boot22:28
markmcclainyeah seems like it would be enough22:29
adam_gesp. since configure is the next state after checkboot22:29
adam_gneed to run out for a bit22:43
drwahlyo, akanda: we had a customer (last year) report that UDP traceroutes fail to floaters22:55
drwahli suspect this is due to the way UDP traceroutes work (semi-random port)22:56
drwahli'm not sure if there is really anything astara should be doing in this case, as it's probably more of a user error, but might be something worth keeping on your radar22:56
*** elopez has quit IRC23:26
openstackgerritMerged openstack/astara: minor changes to orchestrator astara file
markmcclaindrwahl: can file a bug here:

Generated by 2.14.0 by Marius Gedminas - find it at!