*** puranamr has quit IRC | 00:23 | |
*** puranamr has joined #akanda | 00:48 | |
*** puranamr has quit IRC | 00:52 | |
*** puranamr has joined #akanda | 01:41 | |
ryanpetrello | drwahl: so this *is* still an issue if we're having issues w/ load | 01:59 |
---|---|---|
ryanpetrello | meaning if the rug is super backed up | 01:59 |
ryanpetrello | instances that boot won't get DHCP because the rug won't have reconfigured dnsmasq on the router VM yet | 01:59 |
ryanpetrello | I'm all for brainstorming on this, though | 02:00 |
ryanpetrello | in the past we had discussed making dhcp something that could generated on the fly rather than requiring a router push | 02:00 |
ryanpetrello | that could be | 02:00 |
*** cleverdevil has joined #akanda | 02:22 | |
*** puranamr has quit IRC | 02:22 | |
*** cleverdevil has quit IRC | 02:27 | |
*** cleverdevil has joined #akanda | 02:30 | |
*** cleverdevil has quit IRC | 02:33 | |
*** puranamr has joined #akanda | 02:55 | |
*** puranamr has quit IRC | 02:58 | |
*** puranamr has joined #akanda | 03:43 | |
*** sarob has joined #akanda | 03:45 | |
*** puranamr has quit IRC | 03:54 | |
*** puranamr has joined #akanda | 03:59 | |
*** sarob has quit IRC | 04:06 | |
openstackgerrit | Adam Gandelman proposed stackforge/akanda-rug: Get devstack running in the gate (DO NOT MERGE) https://review.openstack.org/187283 | 05:19 |
*** puranamr has quit IRC | 05:20 | |
*** puranamr has joined #akanda | 06:02 | |
*** puranamr has quit IRC | 06:24 | |
*** puranamr has joined #akanda | 13:07 | |
*** puranamr has quit IRC | 13:13 | |
*** puranamr has joined #akanda | 13:14 | |
*** rods1 has joined #akanda | 13:26 | |
*** puranamr has quit IRC | 13:31 | |
*** puranamr has joined #akanda | 13:36 | |
*** puranamr has quit IRC | 13:40 | |
markmcclain | ryanpetrello: yes there are a few ways to solve this.. either make the akanda instance relay DHCP back to the rug or make the DHCP server smarter and fetch on a config miss or make the L2 agent the request | 14:13 |
markmcclain | the last method also ensures that DHCP works on segments without routers attached | 14:13 |
markmcclain | and additionally means that DHCP now shares fate with the HV | 14:14 |
*** puranamr has joined #akanda | 16:18 | |
adam_g | check-functional-dsvm-akandaSUCCESS in 28m 47s | 16:22 |
adam_g | now to figure out how to make the appliance VM boot faster | 16:22 |
ryanpetrello | \o/ | 16:23 |
adam_g | the router VMs boot really slow with non-accelerated qemu | 16:24 |
ryanpetrello | ye | 16:24 |
ryanpetrello | *yep | 16:24 |
adam_g | IIRC its blocking mostly at the pre-mount or post-mount stage of the initrd | 16:24 |
adam_g | i feel like i did something at one point with other DIB images to fix that but dont remember what it was | 16:24 |
adam_g | markmcclain, thoughts on dropping the provider_rules.json stuff? i cant find any evidence of it serving a purpose | 16:40 |
ryanpetrello | (we're not using it) | 16:40 |
adam_g | other than causing https://bugs.launchpad.net/akanda/+bug/1461609 | 16:40 |
openstack | Launchpad bug 1461609 in akanda "if provider rules json is not found, router config fails and the update spins " [Undecided,New] | 16:40 |
*** jordantardif has joined #akanda | 16:41 | |
*** sarob has joined #akanda | 16:45 | |
openstackgerrit | Adam Gandelman proposed stackforge/akanda-rug: Default to empty provider rules config if json not found https://review.openstack.org/188079 | 16:50 |
adam_g | in the meantime.. | 16:50 |
*** puranamr has quit IRC | 17:04 | |
*** puranamr has joined #akanda | 17:04 | |
openstackgerrit | Adam Gandelman proposed stackforge/akanda-rug: Get devstack running in the gate (DO NOT MERGE) https://review.openstack.org/187283 | 17:20 |
openstackgerrit | Adam Gandelman proposed stackforge/akanda-rug: Make blessed an optional dependency https://review.openstack.org/188100 | 17:20 |
openstackgerrit | Adam Gandelman proposed stackforge/akanda-rug: Various devstack cleanups https://review.openstack.org/188101 | 17:20 |
openstackgerrit | Adam Gandelman proposed stackforge/akanda-rug: Get devstack running in the gate (DO NOT MERGE) https://review.openstack.org/187283 | 17:26 |
openstackgerrit | Adam Gandelman proposed stackforge/akanda-rug: Default to empty provider rules config if json not found https://review.openstack.org/188079 | 17:26 |
markmcclain | adam_g: so while DH is not using it.. over time we're going to want to put it back in some form | 17:33 |
markmcclain | adam_g: since the file is unused atm I'm fine dropping it and adding it back when it's actually used | 17:33 |
adam_g | markmcclain, if its on the roadmap to do something with it lets keep it as a placeholder | 17:33 |
adam_g | just to not make it difficult to work in later | 17:34 |
markmcclain | ok it's a feature that some deployers ask about | 17:34 |
puranamr | adam_g, | 17:43 |
puranamr | in ur latest patch, u have a sleep of 340, will it be possible to add a line comment or so, it will be out of the blue for anyone looking later :) | 17:44 |
puranamr | in the run_functional.sh file! | 17:44 |
adam_g | puranamr, yeah, im not planning on merging that. if it ends up we need to keep it, i'll definitely document it | 17:45 |
adam_g | i'd like to see if we can slim down the base appliance image in a way that it boots faster | 17:46 |
puranamr | thanks very much | 17:46 |
puranamr | how much does it take typically today | 17:46 |
puranamr | and where do we want to see it | 17:46 |
adam_g | it takes a few minutes at least currently | 17:47 |
adam_g | somewhere between 240-340sec i think | 17:47 |
adam_g | i wonder if using a single tenant/network/router is okay for our testing purposes | 17:48 |
adam_g | then we only have to boot the thing once | 17:48 |
puranamr | as part of my charter (performance) i have the bootup times also included | 17:51 |
puranamr | what does 45 second SLA for Service VM to come up in DreamCompute mean? | 17:51 |
markmcclain | adam_g: at one point in time I was building an image with buildroot that booted in 7s | 18:14 |
markmcclain | the image was ~ 50meg and contained kernel, busybox, python2.7, tcpdump, iproute(2), nftables | 18:15 |
markmcclain | the debian base is both a curse and a blessing :) | 18:16 |
markmcclain | wondering if something snappy like might be a good compromise | 18:19 |
markmcclain | I do like the updates with rollback side of it | 18:19 |
* ryanpetrello is all for deploying on a smaller base image | 18:34 | |
* ryanpetrello as long as we can add tools to the build we need for debugging when we need to | 18:34 | |
*** puranamr has quit IRC | 18:43 | |
*** puranamr has joined #akanda | 18:49 | |
*** puranamr has quit IRC | 18:51 | |
*** puranamr has joined #akanda | 18:51 | |
openstackgerrit | Merged stackforge/akanda-rug: Make blessed an optional dependency https://review.openstack.org/188100 | 18:52 |
openstackgerrit | Merged stackforge/akanda-rug: Various devstack cleanups https://review.openstack.org/188101 | 18:52 |
openstackgerrit | Merged stackforge/akanda-rug: Default to empty provider rules config if json not found https://review.openstack.org/188079 | 18:52 |
*** puranamr has quit IRC | 19:03 | |
*** anand1712 has joined #akanda | 19:10 | |
*** anand1712 has quit IRC | 19:11 | |
*** puranamr has joined #akanda | 19:22 | |
*** puranamr has quit IRC | 19:32 | |
*** puranamr has joined #akanda | 20:00 | |
adam_g | markmcclain, i wonder if just replacing debian with ubuntu and a modern init system would help? | 20:13 |
adam_g | snappy sounds cool but ihaven't looked at it closely enough to know if its ready for use | 20:13 |
openstackgerrit | Adam Gandelman proposed stackforge/akanda-rug: Get devstack running in the gate (DO NOT MERGE) https://review.openstack.org/187283 | 20:18 |
openstackgerrit | Adam Gandelman proposed stackforge/akanda-rug: devstack: Stop hacking blessed into global-requirements https://review.openstack.org/188170 | 20:18 |
puranamr | markmcclain, for the TII switches do we need QSFP or 20GT? | 20:19 |
*** puranamr has quit IRC | 20:35 | |
*** puranamr has joined #akanda | 20:40 | |
*** puranamr has quit IRC | 20:42 | |
*** puranamr has joined #akanda | 20:44 | |
markmcclain | adam_g: possibly.. jessie should be similarly modern with systemd | 20:49 |
adam_g | right | 20:50 |
markmcclain | the biggest bloat is really the minimal install package set which I think is common for debian and ubuntu | 20:50 |
adam_g | markmcclain, re testing and tempest: with multi-tenant testing, do we want the test tenants to use a shared network/subnet/router or to enable isolation and have each test tenant spin up a new router VM? | 20:51 |
markmcclain | adam_g: ideally I'd like to pass with full isolation | 20:51 |
adam_g | i imagine for functional akanda testing we can just use a single router VM | 20:52 |
markmcclain | we might ahve to hammer on the startup to get there | 20:52 |
markmcclain | puranamr: 20GT? | 20:52 |
* adam_g tries a jessie build | 20:53 | |
markmcclain | it _might_ work if I wrote the ansible stuff right | 20:53 |
puranamr | sorry markmcclain, it was referring to T+ switches | 20:53 |
puranamr | not the TII | 20:53 |
markmcclain | puranamr: I suggested the ly9 since we wouldn't have to mess with QSFP and could recable stuff faster | 20:54 |
puranamr | as agreed earlier, we r in the process of the getting the T3048-LY9 baremetal model | 20:54 |
markmcclain | also for whoever wants to take this... low hanging fruit: http://docs.akanda.io/en/latest/operation.html#installation | 20:55 |
markmcclain | needs to be updated for kilo/liberty ideally with devstack | 20:55 |
puranamr | T+ switch (similar to what DH isusing) | 20:56 |
*** puranamr has quit IRC | 21:00 | |
*** puranamr has joined #akanda | 21:07 | |
openstackgerrit | Adam Gandelman proposed stackforge/akanda-rug: Get devstack running in the gate (DO NOT MERGE) https://review.openstack.org/187283 | 21:21 |
*** rods has joined #akanda | 21:40 | |
*** rods1 has quit IRC | 21:40 | |
*** rods has quit IRC | 21:42 | |
*** rods1 has joined #akanda | 21:42 | |
adam_g | hmm jessie based appliances boot quicker, and they mostly work | 21:58 |
ryanpetrello | \o/ | 22:01 |
openstackgerrit | Adam Gandelman proposed stackforge/akanda-appliance: Restart netfilter-persistent instead of iptables-persistent https://review.openstack.org/188191 | 22:14 |
openstackgerrit | Adam Gandelman proposed stackforge/akanda-rug: Build jessie ramdisk (DO NOT MERGE) https://review.openstack.org/188193 | 22:21 |
openstackgerrit | Adam Gandelman proposed stackforge/akanda-appliance: Allow for building non-wheezy releases https://review.openstack.org/188194 | 22:23 |
markmcclain | adam_g: awesome | 22:47 |
adam_g | testing to see if it helps the gate run | 22:47 |
adam_g | the cross-repo dependency feature in zuul is awesome | 22:47 |
markmcclain | nice | 22:49 |
openstackgerrit | Adam Gandelman proposed stackforge/akanda-rug: Migrate to oslo.utils https://review.openstack.org/183156 | 22:51 |
openstackgerrit | Adam Gandelman proposed stackforge/akanda-rug: Get devstack running in the gate (DO NOT MERGE) https://review.openstack.org/187283 | 23:07 |
openstackgerrit | Adam Gandelman proposed stackforge/akanda-rug: devstack: Stop hacking blessed into global-requirements https://review.openstack.org/188170 | 23:07 |
openstackgerrit | Adam Gandelman proposed stackforge/akanda-rug: Build jessie ramdisk (DO NOT MERGE) https://review.openstack.org/188193 | 23:07 |
*** rods1 has quit IRC | 23:25 | |
adam_g | http://logs.openstack.org/94/188194/1/experimental/check-functional-dsvm-akanda-appliance-src/e1e0c28/logs/screen-ak-rug.txt.gz#_2015-06-03_23_39_42 hmmm | 23:49 |
adam_g | http://logs.openstack.org/94/188194/1/experimental/check-functional-dsvm-akanda-appliance-src/e1e0c28/logs/screen-ak-rug.txt.gz#_2015-06-03_23_37_59 | 23:54 |
adam_g | ryanpetrello, any idea there ^ | 23:54 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!