Delete container replica commands

Learn about the delete container replica commands, its priroty, and the configuration to control it.

The delete container commands are throttled in much the same way as for the replicate container commands. When a delete command is attempted, the current command count is checked and if the datanode is overloaded, another replica is tried or the container is re-queued and attempted again later.

Use the hdds.scm.replication.datanode.delete.container.limit configuration to adjust the limit on the number of delete commands that can be queued on a datanode.

The balancer and delete commands

Unlike the replication commands, there is no priority ordering for delete container replica commands scheduled by the Container Balancer, for several reasons:
  • Delete is less important than replication, as a delayed delete cannot result in data loss.
  • The balancer delete commands are triggered by the completion of a replicate command and this rate of completion naturally throttles the delete.
  • Delete commands are less resource intensive, and hence the datanode should be able to deal with a large number quickly.