kibo node

Inspect nodes, drain them and set overcommit ratios. This page lists the usage, examples, and flags of kibo node and its subcommands, generated from the kibo source code.

Every command on this page also takes the global flags --endpoint, --json, and --timeout. See CLI reference for kibo.

Subcommands:

• kibo node cancel-drain: Withdraw the drain request so the node accepts new work again
• kibo node drain: Request that the node stop accepting new work
• kibo node get: Get one node registration
• kibo node isolate: Declare a node isolated so ctrl releases all its storage and fails its Workspaces
• kibo node list: List node registrations
• kibo node overcommit: Inspect and set the ratios that turn host resources into schedulable CPU and memory
• kibo node retire: Retire a drained or isolated node for good

kibo node cancel-drainCopy Anchor Link

Withdraw the drain request so the node accepts new work again.

See kibo help command-ids for --command-id and --expected-version.

kibo node cancel-drain <node-id> [flags]

kibo node drainCopy Anchor Link

Request that the node stop accepting new work.

See kibo help command-ids for --command-id and --expected-version.

kibo node drain <node-id> [flags]

Examples:

kibo node drain node-a --wait

kibo node getCopy Anchor Link

Get one node registration.

kibo node get <node-id> [flags]

kibo node isolateCopy Anchor Link

Declares that this node can no longer write to any of its storage pools: its host is powered off, or it is cut off from its pools. ctrl cannot verify this claim, so --confirm-host-cannot-write is required: the operator is asserting the host cannot write, and a node that in fact is still writing will corrupt the block files it has open. The request is refused while the node still has a live control session.

ctrl then releases, in one step, the use right of every Volume and retained layer the node held, in every storage pool, shared or not, and every allocation charged to it. Every Workspace whose execution was on the node fails as unavailable and retryable, and commands still addressed to the node fail.

A Workspace whose objects are in a pool another node reaches can be started there; writes still in the old host's page cache may be lost. A Workspace whose objects are in a pool only this node reaches can be started again once the node, or another host with that pool, is back.

See kibo help command-ids for --command-id and --expected-version.

kibo node isolate <node-id> [flags]

kibo node listCopy Anchor Link

Lists node registrations, one row per node. CPU_ALLOC/SCHED and MEM_ALLOC/SCHED are what the node's latest observation reports allocated against what placement admits on it: the host resources times the ratio in effect, whose source CPU_RATIO and MEM_RATIO name.

kibo node list [flags]

Also available as kibo node ls.

kibo node overcommitCopy Anchor Link

Inspect and set the ratios that turn host resources into schedulable CPU and memory.

Subcommands:

• kibo node overcommit get: Show the cluster-wide overcommit ratios
• kibo node overcommit set: Override or restore one node's overcommit ratios
• kibo node overcommit set-default: Replace the cluster-wide overcommit ratios

kibo node overcommit getCopy Anchor Link

Show the cluster-wide overcommit ratios.

kibo node overcommit get [flags]

kibo node overcommit setCopy Anchor Link

Sets this node's own CPU or memory overcommit ratio, which takes the place of the cluster-wide default for that resource. "default" drops the override so the node follows the cluster-wide ratio again. Only the resources given change; the other keeps whatever override it has now. A ratio is a decimal in (0, 16] with at most three fractional digits.

--command-id needs --expected-version and both --cpu and --memory. See kibo help command-ids.

kibo node overcommit set <node-id> [flags]

kibo node overcommit set-defaultCopy Anchor Link

Replaces the cluster-wide CPU and memory overcommit ratios that every node without its own override follows. A node's schedulable CPU and memory are its host resources times the ratio in effect for it; nodes take new values without a restart, and lowering a ratio below what a node already holds only stops new placement there. A ratio is a decimal in (0, 16] with at most three fractional digits. A ratio not given keeps its current value.

The defaults carry their own version, and --command-id needs --expected-version and both ratios. See kibo help command-ids.

kibo node overcommit set-default [flags]

kibo node retireCopy Anchor Link

Ends a node for good: a host that is gone or will not serve again. The node must be DRAINED, or its current incarnation isolated.

The node stops being a member of every storage pool and its pending enrollment grants are revoked. Its registration stays, reading RETIRED, as a record of what it held. It can never register again, and enrolling, draining or isolating the same node ID is refused: a new host takes a new node ID.

A pool only this node reached has no member until another host declares it. Move the disk to another host and declare the pool there under the same ID, and its objects can be started on that host. If the disk is gone too, see kibo storage-pool declare-lost.

See kibo help command-ids for --command-id and --expected-version.

kibo node retire <node-id> [flags]

Examples:

kibo node retire node-a