SREBatchBase is an abstraction to reduce boilerblate in writing cookbooks to reboot clusters or restart daemons of a cluster instead. Currently there are these cookbooks which already use it:
- sre.gitlab.reboot-runner
- sre.ldap.roll-restart-reboot-replica
- sre.k8s.reboot-nodes
- sre.misc-clusters.roll-restart-reboot-docker-registry
- sre.misc-clusters.thumbor.py
- sre.wdqs.restart-nginx
- sre.o11y.roll-restart-reboot-kibana
But the following cookbooks should eventually be converted to use the new framework:
sre.aqs.roll-restart(done)- sre.cassandra.roll-restart
- sre.druid.reboot-workers
- sre.druid.roll-restart-workers
- sre.hadoop.reboot-workers
- sre.hadoop.roll-restart-masters
- sre.hadoop.roll-restart-workers
- sre.kafka.reboot-workers
- sre.kafka.roll-restart-brokers
- sre.kafka.roll-restart-mirror-maker
sre.maps.reboot(done)- sre.mediawiki.restart-appservers
- sre.ores.roll-restart-workers
- sre.presto.reboot-workers
- sre.presto.roll-restart-workers
- sre.wdqs.reboot
- sre.wdqs.restart
- sre.zookeeper.roll-restart-zookeeper
Looking at the names already in use, I think we should also agree on a common naming scheme. It seems to me that
- any reference to "rolling" can go away (it's all quite implicit)
- some cookbooks handle both reboots and daemon restarts at the same, but some have more complex needs (e.g. Hadoop needs only specific daemons restarted)
So maybe a scheme where sre.X.reboot-restart-foodesignates a cookbook which handles both reboots and daemon restarts and sre.X.bar-reboot and sre.X.bar-restart and sre.X.bar-restart-SERVICE are used for services with more specialised needs?