Related-Bug: #1743847
(cherry picked from commit 395d849d3e44f7f58e0f95fccaedc695e835059b)
* Fix file permissions on config cache and unitdata (#94)
(cherry picked from commit 757c1b524f0766b00f9b07d9e6ac118e9f7ce3fa)
* Fix regression in NRPE haproxy check (#95)
These changes fix a regression in the haproxy NRPE checks when
gathering credentials for querying the admin haproxy endpoint,
to determine haproxy backend server health via the admin API.
(cherry picked from commit b2d968cd70a886aa8fee0c21bcd0b2f0501162a0)
* Make HAProxyContext network spaces aware (#92)
Use get_relation_ip in the HAProxy context. get_relation_ip is
IPv6, configuration and network spaces aware.
Partial-Bug: #1735421
(cherry picked from commit 416919061d1ff51d4f8910996856b13151270bf0)
* Support use of HAProxy context in dashboard charm (#98)
The Openstack Dashboard charm only has a public network binding,
which means the default HAProxy behaviour to inspect public,
internal and admin bindings breaks; support override of the
list of bindings to inspect during creation of the context.
(cherry picked from commit a69a671c457f1ca37dd3c26a7d20fd83a813f7cc)
After a config change in which the workload message already matches
what is expected, the units may not yet be idle or worse they are in
an error state and d.wait_for_messages will wait the full timeout
value.
Amulet provides three different checks d.wait checks for idleness,
d.wait_for_status checks for active workload status and
d.wait_for_messages confirms the workload message matches
expectations.
This change adds to the OS _auto_wait_for_status all of these in order
to ensure ready state before inspecting the environment. This change
should also avoid situations where a unit is in error state and amulet
waits the full timeout values before exiting.
Finally, add an _auto_wait_for_status to
keystone_configure_api_version to ensure it is ready to inspect
propagation and address Bug#1668954.