kelvinliu | wallyworld: I can't reproduce the missing sku name even on the 2.6 model upgraded from 2.4 | 01:29 |
---|---|---|
wallyworld | kelvinliu: hmmm, not sure. i thought i reproduced it a while back. maybe ask him in #juju if he's tried it again with 2.7? | 01:36 |
kelvinliu | yep | 01:38 |
kelvinliu | wallyworld: I think I found the problem here, the model has to be upgraded from 2.3 but not 2.4+. | 03:38 |
kelvinliu | wallyworld: im wondering we should land the fix to 2.7 or? | 03:40 |
wallyworld | kelvinliu: 2.8 is fine | 03:41 |
kelvinliu | ok, cool | 03:41 |
wallyworld | i thought it was 2.4, but naybe we chnaged sku type in 2.4 and not 2.5 | 03:42 |
wallyworld | thumper: so yeah, i reproduced the issue where upgrade-charm hook runs "for no reason" deploying a bundle. WTF. need to figure out why | 03:50 |
tlm | ian got 5 min for HO ? | 03:55 |
tlm | wallyworld: | 03:55 |
wallyworld | sure | 03:55 |
thumper | wallyworld: cool, being able to reproduce is one of the hardest issues | 03:55 |
wallyworld | yup | 03:55 |
thumper | timClicks_: we should create a simple doc around agent ratelimiting with examples | 03:57 |
thumper | something to add to the queue | 03:57 |
kelvinliu | wallyworld: it's not related with the SKU change at all. It's because we use unmanaged storage which requires a storage account but we changed to use managed storage for all model after 2.3. so no storage account anymore. | 03:58 |
kelvinliu | for no storage account case, we set the SKU.Name to `Aligned`, but set an empty SKU struct for models do have a storage account (the one upgraded from 2.3) | 04:01 |
wallyworld | kelvinliu: ah i see, nice pickup | 04:18 |
kelvinliu | wallyworld: ah, can i grab u 2mins? | 05:27 |
wallyworld | sure | 05:29 |
flxfoo | Hi there | 09:11 |
flxfoo | I am trying to patch an interface (solr). I have one first issue with current version, calling `add-relation` works fine, but `remove-relation` breaks with ('Unable to determine default scope: no current hook or global scope' | 09:13 |
flxfoo | is that something that is on the interface side or something missing from the charm (which I doubt) ? | 09:13 |
flxfoo | sorry working in the `apache-solr` charm not the interface. | 09:14 |
stickupkid | I guess we use labels on PRs now | 09:48 |
stickupkid | :) | 09:48 |
achilleasa | manadart: on develop, is the core/network.InterfaceInfo populated by the machiner? I want to add a new field to indicate if this is a virtual port (for OVS) | 09:51 |
achilleasa | does that mean that I need to bump the relevant APIs or are they already bumped on develop? | 09:53 |
manadart | achilleasa: Yes, it will be populated by the machiner and instance-poller. I don't think they have been bumped on develop since the 2.8 release, but I may be wrong. | 09:57 |
manadart | achilleasa: Technically it should be bumped right stickupkid? I think there have been instances where we have gotten away with omitempty. | 09:57 |
achilleasa | manadart: stickupkid weren't they bumped when Origin was introduced? | 09:58 |
manadart | achilleasa: Does it matter if it is a virtual port? InterfaceInfo already has IsVirtual returning true if the NIC is VLAN-tagged. | 10:00 |
manadart | achilleasa: Origin went out with 2.8.0. | 10:00 |
achilleasa | manadart: when starting kvm containers you need to pass "<virtualport type='openvswitch' />" to bridge to OVS | 10:01 |
manadart | achilleasa: Righto. | 10:02 |
achilleasa | to get that info in kvm/libvirt I can either add it to NetworkInfo which is decorated as a libvirt.Interface | 10:02 |
achilleasa | or add it to the decorated type as an extra arg | 10:02 |
achilleasa | which means that I need to fetch the OVS bridges and populate that on the fly when I boot the kvm container | 10:02 |
achilleasa | I am doing the latter but thought that maybe if I added this info to NetworkInfo it would be more elegant | 10:03 |
achilleasa | (also my NIC does not use VLANs so I can't use IsVirtual) | 10:04 |
stickupkid | achilleasa, manadart yeah, but I'd need to check | 10:04 |
manadart | achilleasa: We'll need it to flow back and forth right? In to state via InterfaceInfo->NetworkInfo->LinkLayerDevice and back the other way as network config? | 10:04 |
achilleasa | manadart: yes... consirerably more effort than detecting on the fly | 10:04 |
achilleasa | but I think it's better from a modeling perspective | 10:05 |
manadart | achilleasa: Yes. | 10:05 |
achilleasa | given that brctl skips ovs bridges I think it's better to have a more accurate view on the hardware (virtual or not) | 10:05 |
achilleasa | plus, getting it in via the machiner will also allow me to use it in the lxd PR | 10:06 |
* manadart nods. | 10:06 | |
achilleasa | ok, I will do that then and amend my lxd PR afterwards | 10:07 |
achilleasa | we might be able to get away with omitempty here though given that the machiner will report it correctly eventually, right? | 10:07 |
achilleasa | (so no bump would be required?) | 10:08 |
stickupkid | anybody know where this package went in 2.8 https://github.com/juju/juju/tree/2.7/permission | 10:12 |
stickupkid | ah core | 10:12 |
stickupkid | I wish we had a job for `go mod tidy` <-- | 10:14 |
stickupkid | manadart, achilleasa https://github.com/juju/juju/pull/11689 | 10:16 |
achilleasa | stickupkid: I can review this in about 1h if that's OK. Trying to get a fix for the panic for the next 2.7 sha. Can you take a quick look at https://github.com/juju/bundlechanges/pull/62? | 10:27 |
stickupkid | achilleasa, this is super super low priority, yeah I'll take a look | 10:28 |
stickupkid | achilleasa, I love how the documentation for SplitN is wrong "The count determines the number of substrings to return" | 10:31 |
stickupkid | achilleasa, well that's a lie, we asked for 2, you gave us 1 | 10:31 |
achilleasa | stickupkid: it's all best effort :D | 10:32 |
stickupkid | achilleasa, SplitAtMaxNButProbablyWillBeLess() | 10:33 |
manadart | stickupkid achilleasa: Can I get a review of https://github.com/juju/juju/pull/11690 ? It's just regenerated mocks. | 10:58 |
stickupkid | manadart, .... | 11:09 |
stickupkid | ah, you've added 2.7 as well | 11:10 |
flxfoo | endpoint_from_flag() would return an object from a given flag that can be found in the logs. for endpoint_from_name() what would be the name form? | 11:58 |
achilleasa | can I get a CR on https://github.com/juju/juju/pull/11692? | 13:38 |
achilleasa | petevg: can you also take a look at the warning output and let me know if the wording is OK or tweaking is needed? ^^ | 13:39 |
petevg | achilleasa: the warning is a little unclear on the consequences. Will the endpoints show up as dupes, or not show up at all? | 13:43 |
achilleasa | petevg: take a look at the end of the QA steps; basically, you will see them be removed and added | 13:44 |
petevg | achilleasa: got it. I suggested a small change to the wording to reflect that. | 13:46 |
achilleasa | hml: can you take a look at ^ | 14:25 |
hml | achilleasa: i’m knee deep in other stuff… is there a time rush? i can look later | 14:26 |
achilleasa | hml: no but I was hoping to land this some time today so it's included in the next 2.7 sha | 14:26 |
achilleasa | if you can take a look later and hit merge if it's ok that would be great | 14:27 |
hml | achilleasa: will do | 14:28 |
stickupkid | I wish `juju upgrade-controller --build-agent` had a `--splat` flag, which would also update the one in the PATH | 14:39 |
skay | postgresql charm question. I've changed the backup_dir setting, but the DUMP_DIR in the pg_backup_job script still points to the old setting | 21:31 |
skay | shouldn't it change? | 21:32 |
skay | aha. it might be a bug that was fixed and I haven't upgraded the charm | 21:48 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!