* helm: roll master pods when master config changes
The master loads master.toml once at startup (startAdminScripts reads
master.maintenance.scripts and sleep_minutes via viper with no config
watching), and the master ConfigMap is mounted with subPath, which
kubelet never refreshes in a running pod. So a change to
.Values.master.config today updates the ConfigMap but running masters
keep executing the old configuration until something else restarts
them.
Add a checksum/config annotation on the master pod template, following
the existing checksum/s3config pattern on the filer and s3 pods, so a
master config change triggers a rolling restart of the masters.
Signed-off-by: Evans Mungai <mbuevans@gmail.com>
* Guard against duplicate keys
Signed-off-by: Evans Mungai <mbuevans@gmail.com>
* Add checksum to deployment as well
Signed-off-by: Evans Mungai <mbuevans@gmail.com>
* Always ensure the annotation is set
Signed-off-by: Evans Mungai <mbuevans@gmail.com>
* Update comments
Signed-off-by: Evans Mungai <mbuevans@gmail.com>
* Soften stance
Signed-off-by: Evans Mungai <mbuevans@gmail.com>
* helm: merge pod annotations before checksums
---------
Signed-off-by: Evans Mungai <mbuevans@gmail.com>
Co-authored-by: Chris Lu <chris.lu@gmail.com>