lbcli update cluster-config (2.2 and above)

--help

-h

bool

Help for cluster-config.

--parameter/value

string

Parameter: The cluster config parameter name.

Value: The cluster config parameter value to update for the parameter.

• DurationToTurnIntoPermanentFailure - The threshold in seconds for determining that a node has permanently failed. Default value: 4h; minimum value: 1h; maximum value: 256h. Example: 2h0m0s

• VolumeDeletionFromNodeDelay (for internal use). Minimum value: 5s.; maximum value: 2m. Example: 30s

• ClusterName - The cluster name. This should not be empty, and should adhere to the regx name rule.

• ClockDriftIntervalForRaisingEvent - Clock drift events suppression mechanism. Specify the time interval to raise a new event if the clock drift between cluster servers is still detected. Minimum value: 30m; maximum value: 256y. Example: 24h

• DeviceHealthIntervalForRaisingEvent - NVMe SSD failed device events suppression mechanism. Specify the time interval to raise a new event if the NVMe SSD device remains unhealthy. Minimum value: 30m; maximum value: 256y. Example: 24h0m0s

• DefaultQoSName - The default QoS policy of the cluster (to be used when neither volume or project have a QoS policy specified). Example: silver-policy

• AllowedNumRevives - The number of attempts to revive services in specified time window (the default is set to 2. A 0 value will disable the feature). Maximum value: 64.

• RevivesWindowDuration - The time window to monitor the number of auto revive attempts (the default is set to 2 hours). Minimum value: 1h; Maximum value: 1week.

• SpareCapacityReservationPercentage -holds back a percentage of every node's capacity as a buffer, so the cluster keeps free space to rebuild data after a node failure.

  A cluster-admin sets it at runtime with lbcli update cluster-config --parameter=SpareCapacityReservationPercentage --value=<0-50> or through the cluster-config REST API.

  It applies to every node, and every server must run 3.21.1 or later.

  The reservation lowers each node's read-only line, so the node's reported capacity and free space shrink by that percentage. Beyond that, the cluster behaves as it does today, just against the lower line: a node whose data reaches the line goes read-only like a full node, and features that move data (rebalance, eviction, fail-in-place) treat the reserved space as unavailable.

  Lowering the percentage gives the space back, which makes a read-only node writable again and gives a rebuild room after a node failure.

  Nothing warns before a change pushes a node into read-only, and a change is refused while an upgrade, eviction, rebuild or migration is in progress.