tree 0b9cb3bd2bea5b840be8cb6e6c9356a7c8d9ab8c
parent e9c3b50796b36647bf94e66e7b62b5a3c5a51434
author Jan Kundrát <kundrat@cesnet.cz> 1610389449 +0100
committer Jan Kundrát <kundrat@cesnet.cz> 1610405529 +0100

update and change sysrepo persistency

In sysrepo commit c226ab1 (November 2020), some fields in the internal
sysrepo-monitoring module were renamed from pid to cid without a
revision bump of that module. If a post-November sysrepo ever runs on a
system with pre-November YANG module, sysrepo operations fail with this
error:

 Failed to find "cid" as a sibling to "sysrepo-monitoring:datastore".

...and we started hitting just that in our lab. Given that sysrepo
upstream strongly suggests only ever touching the repository content via
`sysrepocfg`, let's make sure that we follow their suggestion and save
data via `sysrepofg -X` and load them via `sysrepocfg -I`.

This required some dance in order to not break Netopeer's
initialization. I split its provisioning setup in two halves, so that it
installs YANG files (and enables all required features) first, then we
can inject the configuration, and then

Right now this means that we regenerate the NETCONF SSH host key during
each and every boot, but that's only a performance penalty since we
immediately overwrite that configuration via the version stored during
last boot. Still, it would be nice to fit that later.

Change-Id: Ib2060ea850d1121d5025183b6c5b17d8962f435b
Bug: https://tree.taiga.io/project/jktjkt-netconf-cli/issue/206
Bug: https://github.com/sysrepo/sysrepo/issues/2280