0.6.0 → 0.7.1 upgrade + workarounds

[aliziel]

Background: Setup is AWS EKS + Nginx + Crunchy PGO, starting from eoAPI v0.6.0. I had been having issues trying to get to v0.7.0, so the release had been through a few upgrades, downgrades, and failed rollbacks. @emmanuelmathot quickly identified the root cause and was kind enough to jump on a fix.

Database permissions were causing database init to hang (#228) and v1.7.1 enabled a successful upgrade. Wanted to report on a couple of the workarounds I needed in case they weren't unique to my upgrade conditions:

1. ✅ Permissions upgrade job completed successfully
2. ❌ pgstac-migrate job failed
   1. Container threw Could not determine path to get from 0.9.6 to 0.9.5
   2. I hadn't modified the default 0.9.5 value, but I did have issues that required a fresh install and the pgstac.migrations timestamp lines up. Some background migrate process during init might have bumped it to 0.9.6
   3. Set pgstacBootstrap.image.tag = v0.9.6 rather than trying to get back down to 0.9.5
3. ✅ pgstac-migrate job succeeded after setting 0.9.6
4. ❌ Upgrade failed with ingress path collision

   Error: UPGRADE FAILED: failed to create resource: admission webhook "validate.nginx.ingress.kubernetes.io" denied the request: host "_" and path "/raster(/|$)(.*)" is already defined in ingress eoapi/nginx-service-ingress-shared-eoapi
   

   1. Deleted eoapi ingress
5. ✅ Upgrade succeeded after deleting eoapi ingress
6. ❌ Further updates fail without previousVersion set
   1. Upgraded using --set, as per guidance, so value wasn't set in-file for subsequent config updates

I'm working with a temporary preprod cluster, so I had more freedom to brute force it. No live traffic and plenty of tuning still going on. Some of the above may not be applicable/feasible for established setups.

[emmanuelmathot]

Thank you @aliziel . Your feedback has been integrated in the updated for 0.7.3 and all issues have been addressed.