*** swest has quit IRC | 01:14 | |
*** swest has joined #zuul | 01:29 | |
*** Goneri has quit IRC | 01:51 | |
*** bhavikdbavishi has joined #zuul | 02:50 | |
*** bhavikdbavishi has quit IRC | 02:54 | |
*** bhavikdbavishi has joined #zuul | 04:01 | |
*** bhavikdbavishi has quit IRC | 04:30 | |
*** evrardjp has quit IRC | 04:37 | |
*** evrardjp has joined #zuul | 04:37 | |
*** bhavikdbavishi has joined #zuul | 05:00 | |
*** bhavikdbavishi1 has joined #zuul | 05:37 | |
*** bhavikdbavishi has quit IRC | 05:39 | |
*** bhavikdbavishi1 is now known as bhavikdbavishi | 05:39 | |
*** evgenyl has quit IRC | 06:13 | |
*** stevthedev has quit IRC | 06:13 | |
*** mugsie has quit IRC | 06:13 | |
*** Open10K8S has quit IRC | 06:13 | |
*** evgenyl has joined #zuul | 06:15 | |
*** stevthedev has joined #zuul | 06:15 | |
*** mugsie has joined #zuul | 06:15 | |
*** Open10K8S has joined #zuul | 06:15 | |
*** bhavikdbavishi has quit IRC | 06:16 | |
*** bhavikdbavishi has joined #zuul | 06:36 | |
*** bhavikdbavishi has quit IRC | 07:25 | |
*** sgw has quit IRC | 08:13 | |
*** bhavikdbavishi has joined #zuul | 08:29 | |
*** evrardjp has quit IRC | 09:01 | |
*** evrardjp has joined #zuul | 09:07 | |
*** bhavikdbavishi has quit IRC | 10:44 | |
*** bhavikdbavishi has joined #zuul | 11:50 | |
*** bhavikdbavishi has quit IRC | 13:31 | |
mordred | fungi: I believe he means the construct in jenkins where you can list a single static node via multiple labels so that it is matched for work when a job requests any of those labels. | 13:45 |
---|---|---|
*** bhavikdbavishi has joined #zuul | 15:50 | |
*** evrardjp has quit IRC | 16:18 | |
*** evrardjp has joined #zuul | 16:24 | |
*** evrardjp has quit IRC | 16:35 | |
*** evrardjp has joined #zuul | 16:35 | |
*** bhavikdbavishi has quit IRC | 16:39 | |
*** sgw has joined #zuul | 17:11 | |
*** bhavikdbavishi has joined #zuul | 17:52 | |
*** bhavikdbavishi has quit IRC | 19:05 | |
*** bhavikdbavishi has joined #zuul | 19:32 | |
*** Goneri has joined #zuul | 19:45 | |
fungi | well, i mean, we labelled non-static nodes too (not that jenkins knows the difference between static and ephemeral nodes) | 19:49 |
fungi | the idea was that you label nodes based on their capabilities and then write the job to require a specific capability | 19:50 |
fungi | setting aside the fact that nodepool generally just builds resources on demand, it's not all that different from the association between node labels and images right? | 19:52 |
clarkb | fungi: one difference is jenkins would do the set union of requirements | 19:53 |
clarkb | nodepool wont combine a x86 and ubuntu and 8vcpu for you | 19:54 |
clarkb | but you can make a label for that if you need it | 19:54 |
clarkb | itsessentially what we've done for highmem, numa, and gpu nodes | 19:54 |
fungi | i did not realize you could have a jenkins job require multiple labels, but that's definitely something nodepool lacks (we essentially have to explode the possible combinations we care about) | 19:57 |
*** bhavikdbavishi has quit IRC | 20:15 | |
jkt | can I tell nodepool from what network to allocate floating ipv4 addresses? | 21:57 |
jkt | (using the openstack driver) | 21:57 |
jkt | it's unfortunately consistently picking public floating ipv4 addresses from a wrong network, even though there's just one neutron router connected to the network the VMs are in | 22:05 |
jkt | digging through the sources, this looks like an openstacksdk issue :( | 22:27 |
*** Goneri has quit IRC | 22:31 | |
*** armstrongs has joined #zuul | 22:39 | |
*** armstrongs has quit IRC | 22:48 | |
clarkb | jkt: you can set it via clouds.yaml | 23:05 |
clarkb | I forget the exact settings thoufh | 23:05 |
* jkt keeps grepping, but this is not exactly a nice keyword to be searching for | 23:22 | |
jkt | clarkb: I tried setting routes_externally, routes_ipv4_externally and default_interface for the network that I would like to use, but that doesn't appear to help | 23:26 |
fungi | i see there's a floating_ip_source option but the documentation doesn't have any info on what all it can do | 23:35 |
fungi | we only ever use it by setting it to None signifying not to try to use floating ips on specific providers | 23:36 |
fungi | but i wonder if you can specify a pool or something there | 23:36 |
* jkt checks, thanks | 23:38 | |
jkt | fungi: that appears to be just for determining whether to use nova-network or neutron | 23:39 |
fungi | ahh, nevermind then | 23:39 |
fungi | mordred's our openstacksdk whisperer, maybe he can think of something | 23:40 |
jkt | mordred: I'm trying to get nodepool working, but our neutron has many public networks, and openstacksdk is picking the wrong one. I've traced this to wait_for_server(auto_ip=True) -> get_active_server -> add_ips_to_server -> _add_auto_ip -> available_floating_ip -> _neutron_available_floating_ips -> _neutron_create_floating_ip | 23:47 |
jkt | mordred: it seems that _neutron_create_floating_ip() is the culprit; no `network` arg is passed, so it just calls to _get_floating_network_id() which returns the first one, which is a wrong choice for me | 23:49 |
jkt | mordred: I have a neutron router between my network to which my VMs are connected; I am hoping for openstacksdk to look at my routers (just one), see that it is connected to both my network and the correct public one, and choose to use the public network for floating ip requests | 23:50 |
jkt | I wonder why our company-internal deities decided to publish a network that I'm not supposed to use, but meh | 23:51 |
jkt | aaargh, and that's exactly what _get_floating_network_id() would do if only get_external_ipv4_floating_networks() did not return a list of valid networks | 23:55 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!