During controller-1 first boot, sysinv-agent starts before
before the IPsec tunnel is established. This causes all NFS mount
attempts for sysinv.conf to fail. The agent
starts without sysinv.conf and operates using oslo_config defaults,
which is functional for RPC and puppet manifest processing. However,
without sysinv.conf the logging configuration is absent and all agent
logs are lost (redirected to /dev/null).
controllerconfig.service runs after sysinv-agent, establishes the
IPsec tunnel via ipsec-client, and mounts /opt/platform. At this
point sysinv.conf becomes accessible over NFS.
Add a restart of sysinv-agent in controller_config after
mount_platform_dir succeeds when sysinv.conf was not obtained at
boot. This is event-driven and ensures the agent reinitialises with
the correct logging configuration without relying on timers or
background retry loops.
Also add retry logic inside mount_platform_dir for the NFS case to
handle the race condition where the IPsec tunnel has just been
established but the NFS service is not yet ready, which previously
caused controllerconfig to exit with fatal_error immediately on the
first mount failure.
extending check_mgmt_conectivity() check and sysinv-agent restart logic
to storage and workerconfig
Test Plan:
PASS Veify sysinv.conf mounted after restart
PASS Verify logs present in sysinv.log in first boot
PASS Verify fresh install/bootstrap AIO-DX,STD(worker+storage),
DC,AIO-SX
PASS Verify no increase in DOR recovery time
Closes-Bug: 2142290
Change-Id: I7c4cabd326e476cc2de41136e5ae525c0f0bfbbd
Signed-off-by: rummadis <ramu.ummadishetty@windriver.com>
d79e32e6c6