IMPORTANT: No additional bug fixes or documentation updates
will be released for this version. For the latest information, see the
current release documentation.
RollingStrategyConfig
editRollingStrategyConfig
editPerforms inline, rolling configuration changes that mutate existing containers. TIP: This is the fastest way to update a plan, but can fail for complex plan changes, such as topology changes. Also, this is less safe for configuration changes that leave a cluster in a non running state. NOTE: When you perform a major version upgrade, and 'group_by' is set to '*all*';, rolling is required.
Properties
edit-
allow_inline_resize(boolean) - Whether we allow changing the capacity of instances (default false). This is currently implemented by stopping, re-creating then starting the affected instance on its associated allocator when performing the changes. NOTES: This requires a round-trip through the allocation infrastructure of the active constructor, as it has to reserve the target capacity without over-committing
-
group_by(string) - Specifies the grouping attribute to use when rolling several instances. Instances that share the same value for the provided attribute key are rolled together as a unit. Examples that make sense to use are '__all__' (roll all instances as a single unit), 'logical_zone_name' (roll instances by zone), '__name__' (roll one instance at a time, the default if not specified). Note that '__all__' is required when performing a major version upgrade
-
shard_init_wait_time(integerasint64) - The time, in seconds, to wait for shards that show no progress of initializing before rolling the next group (default: 10 minutes)
-
skip_synced_flush(boolean) - Whether to skip attempting to do a synced flush on the filesystem of the container (default: false), which is less safe but may be required if the container is unhealthy
Example
edit{
"allow_inline_resize" : true,
"group_by" : "string",
"shard_init_wait_time" : 0,
"skip_synced_flush" : true
}