oracle.kv.impl.admin.plan.MigrateSNPlan Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of oracle-nosql-server Show documentation
Show all versions of oracle-nosql-server Show documentation
NoSQL Database Server - supplies build and runtime support for the server (store) side of the Oracle NoSQL Database.
The newest version!
/*-
* Copyright (C) 2011, 2018 Oracle and/or its affiliates. All rights reserved.
*
* This file was distributed by Oracle as part of a version of Oracle NoSQL
* Database made available at:
*
* http://www.oracle.com/technetwork/database/database-technologies/nosqldb/downloads/index.html
*
* Please see the LICENSE file included in the top-level directory of the
* appropriate version of Oracle NoSQL Database for a copy of the license and
* additional information.
*/
package oracle.kv.impl.admin.plan;
import java.util.HashSet;
import java.util.Set;
import oracle.kv.impl.admin.IllegalCommandException;
import oracle.kv.impl.admin.Admin;
import oracle.kv.impl.admin.param.AdminParams;
import oracle.kv.impl.admin.param.ArbNodeParams;
import oracle.kv.impl.admin.param.Parameters;
import oracle.kv.impl.admin.param.RepNodeParams;
import oracle.kv.impl.admin.param.StorageNodeParams;
import oracle.kv.impl.admin.plan.task.ChangeServiceAddresses;
import oracle.kv.impl.admin.plan.task.ConfirmSNStatus;
import oracle.kv.impl.admin.plan.task.DeployAdmin;
import oracle.kv.impl.admin.plan.task.DeployMultipleRNs;
import oracle.kv.impl.admin.plan.task.MigrateParamsAndTopo;
import oracle.kv.impl.admin.plan.task.NewAdminParameters;
import oracle.kv.impl.admin.plan.task.NewArbNodeParameters;
import oracle.kv.impl.admin.plan.task.NewRepNodeParameters;
import oracle.kv.impl.admin.plan.task.UpdateAdminHelperHost;
import oracle.kv.impl.admin.plan.task.UpdateHelperHostV2;
import oracle.kv.impl.admin.plan.task.VerifyBeforeMigrate;
import oracle.kv.impl.topo.AdminId;
import oracle.kv.impl.topo.ArbNode;
import oracle.kv.impl.topo.Datacenter;
import oracle.kv.impl.topo.RepGroup;
import oracle.kv.impl.topo.RepNode;
import oracle.kv.impl.topo.StorageNodeId;
import oracle.kv.impl.topo.Topology;
import com.sleepycat.persist.model.Persistent;
/**
* A plan for moving all the services from one storage node to another node in
* the KV Store.
*
* We currently require that the destination node does not host any other
* services. This precludes the ability to move services from a defunct storage
* node onto one that is already part of the store. For example, suppose the
* store is currently deployed on storage nodes 1 - 100, and suppose that
* storage node 2 (sn2) fails. The MigrateSNPlan requires that sn2 is replaced
* with a new node, such as sn101. We don't provide the ability to move sn2's
* services to sn1, sn3-100.
*
* If we did permit sn2's services to move to an existing node, the kvstore
* would become imbalanced, in terms of performance. We would also have to
* decide whether we were willing to reduce availability by letting two rep
* nodes from the same group exist on a single storage node. And finally, if we
* did permit this, we would have to provide some way to repair the situation,
* and move these relocated services back to some other sn when one became
* available.
*
* For this reason, it is a conscious decision to avoid supporting that case
* for R1.
*
* R1 note: The user must let the MigrateSNPlan succeed. If it does not, she
* will have to also replace the failed destination. For example, suppose she
* attempts to replace sn1 with sn77. If that plan has some fatal flaw (i.e a
* problem with sn77), the metadata may have been transfered, but the
* deployments will not succeed. In that case, she must cancel the {@literal
* sn1->sn77} plan, and and replace sn77 with the some other sn.
*
* R3 notes:
* - It's now our goal to be able to recover from an incomplete
* MigrateSNPlan. The user should be able to restart sn1 and repair the
* topology, or re-execute the migration (as required above in R1), or migrate
* to a new SN. The latter will probably still require some degree of user
* intervention, by running the RepairPlan.
*
* - It was an implementation goal to be able to use the RelocateRN task, which
* is used by elasticity plans, to move RNs from one SN to another. That would
* reduce duplicate code, but more importantly the RelocateRNTask is more
* atomic and more robust. This turned out to be non-trivial, because the
* RelocateRN task assumes the old SN is up, while in this migrate-sn
* situation, the old SN should be down, as the migrate-sn plan is supposed to
* be used when the hardware is down. This needs investigation in the future.
*
* Migrating an SN between datacenters is not permitted.
*/
@Persistent
public class MigrateSNPlan extends TopologyPlan {
private static final long serialVersionUID = 1L;
private AdminId foundAdminId;
StorageNodeId oldNode;
StorageNodeId newNode;
/*
* httpPort is no longer used, but it is kept here
* for serialization compatibility.
*/
@SuppressWarnings("unused")
@Deprecated
private int httpPort;
/**
* Constructs a plan for replacing a node in the store. Any resources
* known to be allocated to the old node will be moved to the new node.
*
* @param oldNode the node getting replaced
* @param newNode the node to replace the old node
*/
public MigrateSNPlan(String name,
Planner planner,
Topology topology,
StorageNodeId oldNode,
StorageNodeId newNode) {
super(name, planner, topology);
this.oldNode = oldNode;
this.newNode = newNode;
/*
* Check the store for correctness before attempting the migration.
* Hope to stave off compounding failures by fixing/checking before a
* new migration.
*/
addTask(new VerifyBeforeMigrate(this, oldNode, newNode));
/*
* Check that the nodes exist in the admin metadata, and that services
* are distributed as expected.
*/
validate(topology, oldNode, newNode);
/* Confirm that the new node is alive. */
addTask(new ConfirmSNStatus(this,
newNode,
true /* shouldBeRunning*/,
"Please ensure that " + newNode +
" is deployed and running before " +
"attempting a node migration."));
/* Confirm that the old node is dead. */
addTask(new ConfirmSNStatus(this,
oldNode,
false /* shouldBeRunning*/,
"Please ensure that " + oldNode +
" is stopped before attempting a node " +
"migration."));
/*
* Update params and topology so all services on the old node are moved
* to the new node,and push topology changes to all nodes.
*/
addTask(new MigrateParamsAndTopo(this, oldNode, newNode));
/* Update the membership address held in the target rep groups.*/
addTask(new ChangeServiceAddresses(this, oldNode, newNode));
/* Create the Admin service, if needed */
if (foundAdminId != null) {
addTask(new DeployAdmin(this, newNode, foundAdminId));
// TODO: May need to check admin state after deployment
}
/* Create the RepNode services */
addTask(new DeployMultipleRNs(this, newNode));
/*
* Add a set of tasks at the end to rewrite the peers of the target SNs
* repNodeParams with the full set of helper hosts for the group, to
* improve the robustness of the group when nodes fail. No need to
* bounce the RepNodes -- helper hosts are mutable.
*/
Set affectedRGs = new HashSet();
for (RepGroup rg: topology.getRepGroupMap().getAll()) {
for (RepNode rn : rg.getRepNodes()) {
if (rn.getStorageNodeId().equals(oldNode)) {
affectedRGs.add(rg);
break;
}
}
for (ArbNode an : rg.getArbNodes()) {
if (an.getStorageNodeId().equals(oldNode)) {
affectedRGs.add(rg);
break;
}
}
}
for (RepGroup rg : affectedRGs) {
for (RepNode rn : rg.getRepNodes()) {
if (!rn.getStorageNodeId().equals(oldNode)) {
addTask(new UpdateHelperHostV2(this,
rn.getResourceId(),
rg.getResourceId()));
addTask(new NewRepNodeParameters(this,
rn.getResourceId()));
}
}
for (ArbNode an : rg.getArbNodes()) {
if (!an.getStorageNodeId().equals(oldNode)) {
addTask(new UpdateHelperHostV2(this,
an.getResourceId(),
rg.getResourceId()));
addTask(new NewArbNodeParameters(this,
an.getResourceId()));
}
}
}
if (foundAdminId != null) {
/* Update the other Admins' helper host addresses. */
Admin admin = planner.getAdmin();
Parameters parameters = admin.getCurrentParameters();
for (AdminParams ap : parameters.getAdminParams()) {
AdminId aid = ap.getAdminId();
/* No need to update the migrated Admin. */
if (aid.equals(foundAdminId)) {
continue;
}
StorageNodeParams snp = parameters.get(ap.getStorageNodeId());
String hostname = snp.getHostname();
int registryPort = snp.getRegistryPort();
addTask(new UpdateAdminHelperHost(this, aid));
addTask(new NewAdminParameters
(this, hostname, registryPort, aid));
}
}
}
private void validate(Topology topology,
StorageNodeId oldNode1,
StorageNodeId newNode1) {
Parameters parameters = planner.getAdmin().getCurrentParameters();
/* Confirm that the old node exists in the params and topology. */
if (topology.get(oldNode1) == null) {
throw new IllegalCommandException
(oldNode1 + " does not exist in the topology and cannot " +
"be migrated");
}
if (parameters.get(oldNode1) == null) {
throw new IllegalCommandException
(oldNode1 + " does not exist in the parameters and cannot " +
"be migrated");
}
/* Confirm that the new node exists in the params and topology. */
if (topology.get(newNode1) == null) {
throw new IllegalCommandException
(newNode1 + " is not in the topology and must be deployed " +
"before migrating " + oldNode1);
}
if (parameters.get(newNode1) == null) {
throw new IllegalCommandException
(newNode1 + " is not in the parameters and must be deployed " +
"before migrating " + oldNode1);
}
/*
* Services should not exist in the topology and parameters for both
* the new node and the old node.
*
* If the plan has never been run, there should be no services on the
* new node. If the plan has already been run, all services should have
* been moved to the new node, and there should be nothing on the old
* node, according to the topology and params.
*
* We permit running the plan though, in case the plan had
* not completed successfully. It may be that the metadata was
* changed and committed, but the actual deploy actions had not taken
* effect.
*/
boolean oldNodeAdminsExist = false;
boolean newNodeAdminsExist = false;
boolean oldNodeRNsExist = false;
boolean newNodeRNsExist = false;
boolean oldNodeANsExist = false;
boolean newNodeANsExist = false;
for (AdminParams ap: parameters.getAdminParams()) {
if (ap.getStorageNodeId().equals(oldNode1)) {
oldNodeAdminsExist = true;
foundAdminId = ap.getAdminId();
}
if (ap.getStorageNodeId().equals(newNode1)) {
newNodeAdminsExist = true;
foundAdminId = ap.getAdminId();
}
}
for (RepNodeParams rnp: parameters.getRepNodeParams()) {
if (rnp.getStorageNodeId().equals(oldNode1)) {
oldNodeRNsExist = true;
}
if (rnp.getStorageNodeId().equals(newNode1)) {
newNodeRNsExist = true;
}
}
for (ArbNodeParams anp: parameters.getArbNodeParams()) {
if (anp.getStorageNodeId().equals(oldNode1)) {
oldNodeANsExist = true;
}
if (anp.getStorageNodeId().equals(newNode1)) {
newNodeANsExist = true;
}
}
if (oldNodeRNsExist && newNodeRNsExist) {
throw new IllegalCommandException
("Cannot migrate services from " + oldNode1 + " to " +
newNode1 + " because " + newNode1 + " is already in use");
}
if (oldNodeAdminsExist && newNodeAdminsExist) {
throw new IllegalCommandException
("Cannot migrate services from " + oldNode1 + " to " +
newNode1 + " because " + newNode1 + " is already in use");
}
if (oldNodeANsExist && newNodeANsExist) {
throw new IllegalCommandException
("Cannot migrate services from " + oldNode1 + " to " +
newNode1 + " because " + newNode1 + " is already in use");
}
if ((!oldNodeAdminsExist) &&
(!oldNodeRNsExist) &&
(!oldNodeANsExist) &&
(!newNodeAdminsExist) &&
(!newNodeRNsExist) &&
(!newNodeANsExist)) {
throw new IllegalCommandException
("No services on " + oldNode1 + " or " + newNode1 +
", nothing to migrate");
}
final Datacenter oldDC = topology.getDatacenter(oldNode1);
final Datacenter newDC = topology.getDatacenter(newNode1);
if (!oldDC.equals(newDC)) {
throw new IllegalCommandException(
"Cannot migrate services from " + oldNode1 + " to " +
newNode1 + " because they belong to different zones." +
" Node " + oldNode1 + " belongs to zone: " + oldDC +
". Node " + newNode1 + " belongs to zone: " + newDC +
".");
}
}
/*
* No-arg ctor for use by DPL.
*/
@SuppressWarnings("unused")
private MigrateSNPlan() {
}
public StorageNodeId getOldNode() {
return oldNode;
}
public StorageNodeId getNewNode() {
return newNode;
}
@Override
void preExecutionSave() {
/* Nothing to do. */
}
@Override
public String getDefaultName() {
return "Migrate Storage Node";
}
}
© 2015 - 2025 Weber Informatics LLC | Privacy Policy