| kklimonda | hmm, when upgrading magnum from one version to another, and them upgrading cluster deployment, heat stack is not updated with new template | 10:20 |
|---|---|---|
| kklimonda | is it by design, or is there a bug somewhere? | 10:21 |
| brtknr | kklimonda: this is by design because each cluster uses instance of the heat stack templates available during cluster creation | 12:13 |
| brtknr | kklimonda: since new templates may not be backward compatible | 12:13 |
| kklimonda | right, it seems to be a problem when you want to have long term support for deployed clusters | 12:17 |
| kklimonda | I've noticed a bug in a template, but now I have no way of fixing it in an existing cluster | 12:18 |
| kklimonda | that's also a problem with most labels being read-only post creation - I have deployed cluster with default hyperkube location and now I'm seemingly stuck with v1.18.6 | 12:18 |
| kklimonda | I see an argument about new templates not being backwards compatible, but that seems to be a nightmare for production and maintenance :( | 12:19 |
| kklimonda | for example, If I understand https://review.opendev.org/c/openstack/magnum/+/799040 correctly, it seems currently deployed clusters with wallaby can't be upgraded to 1.21+ at all | 12:22 |
| brtknr | kklimonda: you can update container_infra_prefix variable in the heat stack directly with --existing flag | 12:28 |
| brtknr | as long as your etcd_volume is enabled in the cluster | 12:28 |
| brtknr | i'd recommend testing this on a discard cluster first though | 12:28 |
| brtknr | i.e the master nodes need to have etcd_volumes attached as cinder volumes | 12:29 |
| kklimonda | because running in-place upgrade will redeploy master nodes? | 12:36 |
| brtknr | kklimonda: thats right and if its not using external etcd volume, the state of the cluster gets lost | 12:39 |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!