The CLIENT KILL command closes a given client connection. Up to Redis 2.8.11 it was possible to close a connection only by client address, using the following form:
CLIENT KILL addr:port
ip:port should match a line returned by the CLIENT LIST command (
However starting with Redis 2.8.12 or greater, the command accepts the following form:
CLIENT KILL <filter> <value> ... ... <filter> <value>
With the new form it is possible to kill clients by different attributes instead of killing just by address. The following filters are available:
CLIENT KILL ADDR ip:port. This is exactly the same as the old three-arguments behavior.
CLIENT KILL ID client-id. Allows to kill a client by its unique
IDfield, which was introduced in the CLIENT LIST command starting from Redis 2.8.12.
CLIENT KILL TYPE type, where type is one of
mastertype is available from v3.2). This closes the connections of all the clients in the specified class. Note that clients blocked into the MONITOR command are considered to belong to the
CLIENT KILL SKIPME yes/no. By default this option is set to
yes, that is, the client calling the command will not get killed, however setting this option to
nowill have the effect of also killing the client calling the command.
It is possible to provide multiple filters at the same time. The command will handle multiple filters via logical AND. For example:
CLIENT KILL addr 127.0.0.1:6379 type slave
is valid and will kill only a slaves with the specified address. This format containing multiple filters is rarely useful currently.
When the new form is used the command no longer returns
OK or an error, but instead the number of killed clients, that may be zero.
*CLIENT KILL and Redis Sentinel
Recent versions of Redis Sentinel (Redis 2.8.12 or greater) use CLIENT KILL in order to kill clients when an instance is reconfigured, in order to force clients to perform the handshake with one Sentinel again and update its configuration.
Due to the single-threaded nature of Redis, it is not possible to kill a client connection while it is executing a command. From the client point of view, the connection can never be closed in the middle of the execution of a command. However, the client will notice the connection has been closed only when the next command is sent (and results in network error).
When called with the three arguments format:
Simple string reply:
OK if the connection exists and has been closed
When called with the filter / value format:
Integer reply: the number of clients killed.