| devananda | self.assertLess(len(delta), 1000*0.1) | 00:00 |
|---|---|---|
| devananda | but the check is for # moved < 100 | 00:00 |
| devananda | only 10 nodes should be assigned to the new conductor, so of course that passes :) | 00:01 |
| devananda | but actualy about 20 moved | 00:01 |
| lifeless | 10 is too strict though | 00:01 |
| lifeless | I agree that 100 is too high | 00:01 |
| lifeless | + # Adding 1 conductor to a set of 10 should move 1 in 11 nodes, or 9% | 00:02 |
| lifeless | + # of the ring - using 10% as an upper threshold allows for some hash | 00:02 |
| lifeless | the that should say 1 to a set of 100 | 00:02 |
| devananda | it is consistently moving twice that, though | 00:02 |
| lifeless | (and 1 in 101) | 00:03 |
| devananda | so 0.9% ? | 00:03 |
| lifeless | yeah | 00:03 |
| lifeless | (I mean, its still not 20 :)) | 00:03 |
| lifeless | the count should be nodes*upper_bound_moved, or 1000*0.01 | 00:03 |
| devananda | right - so, the "extra zero" aside, it is moving 2x what it theoretically should | 00:04 |
| devananda | which may be much better than it was previously :) | 00:04 |
| lifeless | that migt be double counting | 00:04 |
| devananda | ooh | 00:04 |
| lifeless | oh its hugely better | 00:04 |
| lifeless | but I'm not sure its double counting yet, sec/. | 00:04 |
| devananda | looks like its not | 00:05 |
| lifeless | so no, its not double counting AFAICT | 00:05 |
| lifeless | so, I'd rather you didn't use uuids in this test | 00:06 |
| lifeless | since the output is easier to inspect as it stands IMO | 00:06 |
| lifeless | up to you of course | 00:06 |
| devananda | sure | 00:07 |
| lifeless | also for some reason I didn't get a poll invite for ironic | 00:07 |
| devananda | actual randomness in a unit test is counterproductive | 00:07 |
| devananda | perhaps no one else ran? | 00:07 |
| lifeless | ah | 00:07 |
| lifeless | that would asplain it | 00:07 |
| lifeless | ok | 00:07 |
| lifeless | ah | 00:10 |
| lifeless | I think I know | 00:10 |
| lifeless | this is the line | 00:10 |
| lifeless | + for p in range((2 ** CONF.hash_partition_exponent) / len(hosts)): | 00:10 |
| lifeless | causing the behaviour you are seeing. | 00:10 |
| lifeless | devananda: your push on the 27th was a no-op rebase? | 00:15 |
| devananda | yes. about to push again. few more min | 00:15 |
| lifeless | devananda: so I have a fix | 00:15 |
| lifeless | I can just do the changes we discussed to the test and push it up | 00:15 |
| lifeless | devananda: if you like | 00:15 |
| devananda | nope, please wait | 00:15 |
| devananda | i have a bunch of other things in there too | 00:16 |
| lifeless | k | 00:16 |
| devananda | i dont have the fix for the *2 yet | 00:17 |
| lifeless | I'll add it in once you push | 00:17 |
| devananda | k | 00:18 |
| devananda | this also has pep8 fixes and whatnot | 00:18 |
| lifeless | shrug :) | 00:18 |
| openstackgerrit | Devananda van der Veen proposed a change to openstack/ironic: Fix hash ring stability https://review.openstack.org/118932 | 00:18 |
| openstackgerrit | Devananda van der Veen proposed a change to openstack/ironic: Add "active" tracking to nodes and conductors https://review.openstack.org/124493 | 00:18 |
| devananda | and a conductor-leaves-the-ring test too | 00:18 |
| devananda | bbiab | 00:20 |
| lifeless | argh f*king easy_install | 00:34 |
| lifeless | cryptography has six as a setup_requires | 00:36 |
| lifeless | so we need six earlier than that | 00:36 |
| *** pcrews has joined #openstack-ironic | 01:09 | |
| *** pcrews has quit IRC | 01:16 | |
| openstackgerrit | lifeless proposed a change to openstack/ironic: Fix hash ring stability https://review.openstack.org/118932 | 01:36 |
| lifeless | devananda: ^ | 01:36 |
| devananda | lifeless: ack. pulling it into my tree | 01:44 |
| *** jrist has quit IRC | 01:45 | |
| devananda | lifeless: it still has my "*2" hack in the tests. is that intentional? | 01:45 |
| openstackgerrit | Devananda van der Veen proposed a change to openstack/ironic: Add "active" tracking to nodes and conductors https://review.openstack.org/124493 | 01:46 |
| openstackgerrit | Devananda van der Veen proposed a change to openstack/ironic: WIP: add periodic task to rebuild hash rings https://review.openstack.org/124610 | 01:46 |
| devananda | also, first poke at implementing the rebalance periodic task ^ | 01:47 |
| devananda | afk for the evening now | 01:48 |
| lifeless | oh | 01:50 |
| lifeless | I didn't realise you committed workaround code in the tests | 01:50 |
| lifeless | devananda: your path for removes is wrong :) | 01:51 |
| lifeless | devananda: it has to move 1/N | 01:51 |
| lifeless | devananda: because thats the # on the node that is going away | 01:51 |
| lifeless | devananda: hack removed | 01:53 |
| openstackgerrit | lifeless proposed a change to openstack/ironic: Fix hash ring stability https://review.openstack.org/118932 | 01:53 |
| openstackgerrit | lifeless proposed a change to openstack/ironic: Fix hash ring stability https://review.openstack.org/118932 | 02:14 |
| lifeless | devananda: ^ all finished pushing stuff now | 02:14 |
| *** yarker has quit IRC | 04:48 | |
| *** yarker has joined #openstack-ironic | 04:53 | |
| *** rushiagr_away is now known as rushiagr | 06:00 | |
| *** rushiagr is now known as rushiagr_away | 06:10 | |
| *** rushiagr_away is now known as rushiagr | 06:12 | |
| *** shakamunyi has joined #openstack-ironic | 06:37 | |
| *** rushiagr is now known as rushiagr_away | 06:42 | |
| *** pensu has joined #openstack-ironic | 07:16 | |
| *** igordcard has joined #openstack-ironic | 07:38 | |
| *** rushiagr_away is now known as rushiagr | 07:54 | |
| *** rushiagr is now known as rushiagr_away | 07:56 | |
| *** romcheg has joined #openstack-ironic | 08:25 | |
| *** andreykurilin_ has joined #openstack-ironic | 08:55 | |
| *** romcheg has quit IRC | 09:17 | |
| *** shakamunyi has quit IRC | 09:49 | |
| *** andreykurilin_ has quit IRC | 09:55 | |
| *** lazy_prince has quit IRC | 11:16 | |
| *** jroll has quit IRC | 11:41 | |
| *** rushiagr_away is now known as rushiagr | 11:44 | |
| *** jroll has joined #openstack-ironic | 11:46 | |
| *** romcheg has joined #openstack-ironic | 11:57 | |
| *** romcheg has quit IRC | 11:59 | |
| *** romcheg has joined #openstack-ironic | 12:01 | |
| *** killer_prince has joined #openstack-ironic | 12:02 | |
| *** killer_prince is now known as lazy_prince | 12:02 | |
| *** romcheg has quit IRC | 12:02 | |
| *** lazy_prince has quit IRC | 12:30 | |
| *** rushiagr is now known as rushiagr_away | 12:33 | |
| *** russell_h has quit IRC | 13:31 | |
| *** pensu has quit IRC | 13:40 | |
| *** rushiagr_away is now known as rushiagr | 13:52 | |
| *** pensu has joined #openstack-ironic | 13:56 | |
| *** russellb has joined #openstack-ironic | 14:02 | |
| *** rushiagr is now known as rushiagr_away | 14:03 | |
| *** rushiagr_away is now known as rushiagr | 14:06 | |
| *** romcheg has joined #openstack-ironic | 14:13 | |
| *** rushiagr is now known as rushiagr_away | 14:33 | |
| *** rushiagr_away is now known as rushiagr | 14:36 | |
| *** romcheg has quit IRC | 14:58 | |
| *** JoshNang has quit IRC | 14:59 | |
| *** romcheg has joined #openstack-ironic | 15:14 | |
| *** krtaylor has quit IRC | 15:18 | |
| *** russell_h has joined #openstack-ironic | 15:25 | |
| *** russell_h has quit IRC | 15:25 | |
| *** russell_h has joined #openstack-ironic | 15:25 | |
| *** shakamunyi has joined #openstack-ironic | 15:48 | |
| *** shakamunyi has quit IRC | 15:53 | |
| *** pcrews has joined #openstack-ironic | 15:56 | |
| *** pcrews has quit IRC | 16:06 | |
| *** romcheg has quit IRC | 16:32 | |
| *** jroll has quit IRC | 16:45 | |
| *** jroll has joined #openstack-ironic | 16:45 | |
| *** shakamunyi has joined #openstack-ironic | 16:50 | |
| *** pcrews has joined #openstack-ironic | 17:27 | |
| *** andreykurilin_ has joined #openstack-ironic | 18:05 | |
| *** killer_prince has joined #openstack-ironic | 18:18 | |
| *** killer_prince is now known as lazy_prince | 18:18 | |
| *** lazy_prince has quit IRC | 18:22 | |
| *** killer_prince has joined #openstack-ironic | 18:23 | |
| *** killer_prince is now known as lazy_prince | 18:23 | |
| *** lazy_prince has quit IRC | 18:42 | |
| *** killer_prince has joined #openstack-ironic | 18:44 | |
| *** killer_prince is now known as lazy_prince | 18:44 | |
| *** igordcard has quit IRC | 18:45 | |
| *** rushiagr is now known as rushiagr_away | 19:04 | |
| *** rushiagr_away is now known as rushiagr | 19:14 | |
| *** mgagne has quit IRC | 19:17 | |
| *** hemna has joined #openstack-ironic | 19:23 | |
| *** mgagne has joined #openstack-ironic | 19:34 | |
| *** mgagne is now known as Guest73730 | 19:34 | |
| *** krtaylor has joined #openstack-ironic | 19:40 | |
| *** rushiagr is now known as rushiagr_away | 19:50 | |
| *** pensu has quit IRC | 19:52 | |
| *** adam_g has quit IRC | 19:55 | |
| *** adam_g has joined #openstack-ironic | 20:07 | |
| *** adam_g has quit IRC | 20:07 | |
| *** adam_g has joined #openstack-ironic | 20:07 | |
| *** enikanorov has joined #openstack-ironic | 20:08 | |
| lifeless | devananda: had a chance to look ? | 20:10 |
| *** JayF has quit IRC | 20:13 | |
| *** aweeks has joined #openstack-ironic | 20:18 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!