When developing and testing upgrade steps for address space IDs, I achieved an error state during model-cache worker start-up. This is seen many times per second:
2019-10-04 14:07:58 INFO juju.state multiwatcher.go:242 store manager loop failed: error loading entities for model 0d1ea901-aad1-4258-88ea-71269febfe4d: failed to initialise backing for machines:0d1ea901-aad1-4258-88ea-71269febfe4d:0: retrieving space for ID "1": space id "1" not found
2019-10-04 14:07:59 ERROR juju.worker.modelcache worker.go:195 watcher error: space id "1" not found, getting new watcher
In fact it happens in such a torrent, that it rolls logs before the error preamble can be observed.
We should have a back-off in the model-cache for getting a new all-watcher, so that we avoid the log spamming.
https:/