You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
consul/agent/intentions_endpoint.go

607 lines
18 KiB

package agent
import (
"fmt"
"net/http"
"strings"
cachetype "github.com/hashicorp/consul/agent/cache-types"
"github.com/hashicorp/consul/agent/consul"
"github.com/hashicorp/consul/agent/structs"
)
// /v1/connect/intentions
func (s *HTTPHandlers) IntentionEndpoint(resp http.ResponseWriter, req *http.Request) (interface{}, error) {
switch req.Method {
case "GET":
return s.IntentionList(resp, req)
case "POST":
return s.IntentionCreate(resp, req)
default:
return nil, MethodNotAllowedError{req.Method, []string{"GET", "POST"}}
}
}
// GET /v1/connect/intentions
func (s *HTTPHandlers) IntentionList(resp http.ResponseWriter, req *http.Request) (interface{}, error) {
// Method is tested in IntentionEndpoint
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
4 years ago
var args structs.IntentionListRequest
if done := s.parse(resp, req, &args.Datacenter, &args.QueryOptions); done {
return nil, nil
}
if err := s.parseEntMeta(req, &args.EnterpriseMeta); err != nil {
return nil, err
}
var reply structs.IndexedIntentions
defer setMeta(resp, &reply.QueryMeta)
if err := s.agent.RPC("Intention.List", &args, &reply); err != nil {
return nil, err
}
return reply.Intentions, nil
}
// IntentionCreate is used to create legacy intentions.
// Deprecated: use IntentionPutExact.
func (s *HTTPHandlers) IntentionCreate(resp http.ResponseWriter, req *http.Request) (interface{}, error) {
// Method is tested in IntentionEndpoint
var entMeta structs.EnterpriseMeta
if err := s.parseEntMetaNoWildcard(req, &entMeta); err != nil {
return nil, err
}
if entMeta.PartitionOrDefault() != structs.PartitionOrDefault("") {
return nil, BadRequestError{Reason: "Cannot use a partition with this endpoint"}
}
args := structs.IntentionRequest{
Op: structs.IntentionOpCreate,
}
s.parseDC(req, &args.Datacenter)
s.parseToken(req, &args.Token)
if err := decodeBody(req.Body, &args.Intention); err != nil {
return nil, fmt.Errorf("Failed to decode request body: %s", err)
}
if args.Intention.DestinationPartition != "" && args.Intention.DestinationPartition != "default" {
return nil, BadRequestError{Reason: "Cannot specify a destination partition with this endpoint"}
}
if args.Intention.SourcePartition != "" && args.Intention.SourcePartition != "default" {
return nil, BadRequestError{Reason: "Cannot specify a source partition with this endpoint"}
}
args.Intention.FillPartitionAndNamespace(&entMeta, false)
if err := s.validateEnterpriseIntention(args.Intention); err != nil {
return nil, err
}
var reply string
if err := s.agent.RPC("Intention.Apply", &args, &reply); err != nil {
return nil, err
}
return intentionCreateResponse{reply}, nil
}
func (s *HTTPHandlers) validateEnterpriseIntention(ixn *structs.Intention) error {
if err := s.validateEnterpriseIntentionPartition("SourcePartition", ixn.SourcePartition); err != nil {
return err
}
if err := s.validateEnterpriseIntentionPartition("DestinationPartition", ixn.DestinationPartition); err != nil {
return err
}
if err := s.validateEnterpriseIntentionNamespace("SourceNS", ixn.SourceNS, true); err != nil {
return err
}
if err := s.validateEnterpriseIntentionNamespace("DestinationNS", ixn.DestinationNS, true); err != nil {
return err
}
return nil
}
// GET /v1/connect/intentions/match
func (s *HTTPHandlers) IntentionMatch(resp http.ResponseWriter, req *http.Request) (interface{}, error) {
// Prepare args
args := &structs.IntentionQueryRequest{Match: &structs.IntentionQueryMatch{}}
if done := s.parse(resp, req, &args.Datacenter, &args.QueryOptions); done {
return nil, nil
}
var entMeta structs.EnterpriseMeta
if err := s.parseEntMetaNoWildcard(req, &entMeta); err != nil {
return nil, err
}
q := req.URL.Query()
// Extract the "by" query parameter
if by, ok := q["by"]; !ok || len(by) != 1 {
return nil, fmt.Errorf("required query parameter 'by' not set")
} else {
switch v := structs.IntentionMatchType(by[0]); v {
case structs.IntentionMatchSource, structs.IntentionMatchDestination:
args.Match.Type = v
default:
return nil, fmt.Errorf("'by' parameter must be one of 'source' or 'destination'")
}
}
// Extract all the match names
names, ok := q["name"]
if !ok || len(names) == 0 {
return nil, fmt.Errorf("required query parameter 'name' not set")
}
// Build the entries in order. The order matters since that is the
// order of the returned responses.
args.Match.Entries = make([]structs.IntentionMatchEntry, len(names))
for i, n := range names {
ap, ns, name, err := parseIntentionStringComponent(n, &entMeta)
if err != nil {
return nil, fmt.Errorf("name %q is invalid: %s", n, err)
}
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
4 years ago
args.Match.Entries[i] = structs.IntentionMatchEntry{
Partition: ap,
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
4 years ago
Namespace: ns,
Name: name,
}
}
// Make the RPC request
var out structs.IndexedIntentionMatches
defer setMeta(resp, &out.QueryMeta)
if s.agent.config.HTTPUseCache && args.QueryOptions.UseCache {
raw, m, err := s.agent.cache.Get(req.Context(), cachetype.IntentionMatchName, args)
if err != nil {
return nil, err
}
defer setCacheMeta(resp, &m)
reply, ok := raw.(*structs.IndexedIntentionMatches)
if !ok {
// This should never happen, but we want to protect against panics
return nil, fmt.Errorf("internal error: response type not correct")
}
out = *reply
} else {
RETRY_ONCE:
if err := s.agent.RPC("Intention.Match", args, &out); err != nil {
return nil, err
}
if args.QueryOptions.AllowStale && args.MaxStaleDuration > 0 && args.MaxStaleDuration < out.LastContact {
args.AllowStale = false
args.MaxStaleDuration = 0
goto RETRY_ONCE
}
}
out.ConsistencyLevel = args.QueryOptions.ConsistencyLevel()
// We must have an identical count of matches
if len(out.Matches) != len(names) {
return nil, fmt.Errorf("internal error: match response count didn't match input count")
}
// Use empty list instead of nil.
response := make(map[string]structs.Intentions)
for i, ixns := range out.Matches {
response[names[i]] = ixns
}
return response, nil
}
// GET /v1/connect/intentions/check
func (s *HTTPHandlers) IntentionCheck(resp http.ResponseWriter, req *http.Request) (interface{}, error) {
// Prepare args
args := &structs.IntentionQueryRequest{Check: &structs.IntentionQueryCheck{}}
if done := s.parse(resp, req, &args.Datacenter, &args.QueryOptions); done {
return nil, nil
}
var entMeta structs.EnterpriseMeta
if err := s.parseEntMetaNoWildcard(req, &entMeta); err != nil {
return nil, err
}
q := req.URL.Query()
// Set the source type if set
args.Check.SourceType = structs.IntentionSourceConsul
if sourceType, ok := q["source-type"]; ok && len(sourceType) > 0 {
args.Check.SourceType = structs.IntentionSourceType(sourceType[0])
}
// Extract the source/destination
source, ok := q["source"]
if !ok || len(source) != 1 {
return nil, fmt.Errorf("required query parameter 'source' not set")
}
destination, ok := q["destination"]
if !ok || len(destination) != 1 {
return nil, fmt.Errorf("required query parameter 'destination' not set")
}
// We parse them the same way as matches to extract partition/namespace/name
args.Check.SourceName = source[0]
if args.Check.SourceType == structs.IntentionSourceConsul {
ap, ns, name, err := parseIntentionStringComponent(source[0], &entMeta)
if err != nil {
return nil, fmt.Errorf("source %q is invalid: %s", source[0], err)
}
args.Check.SourcePartition = ap
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
4 years ago
args.Check.SourceNS = ns
args.Check.SourceName = name
}
// The destination is always in the Consul format
ap, ns, name, err := parseIntentionStringComponent(destination[0], &entMeta)
if err != nil {
return nil, fmt.Errorf("destination %q is invalid: %s", destination[0], err)
}
args.Check.DestinationPartition = ap
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
4 years ago
args.Check.DestinationNS = ns
args.Check.DestinationName = name
var reply structs.IntentionQueryCheckResponse
if err := s.agent.RPC("Intention.Check", args, &reply); err != nil {
return nil, err
}
return &reply, nil
}
// IntentionExact handles the endpoint for /v1/connect/intentions/exact
func (s *HTTPHandlers) IntentionExact(resp http.ResponseWriter, req *http.Request) (interface{}, error) {
switch req.Method {
case "GET":
return s.IntentionGetExact(resp, req)
case "PUT":
return s.IntentionPutExact(resp, req)
case "DELETE":
return s.IntentionDeleteExact(resp, req)
default:
return nil, MethodNotAllowedError{req.Method, []string{"GET", "PUT", "DELETE"}}
}
}
// GET /v1/connect/intentions/exact
func (s *HTTPHandlers) IntentionGetExact(resp http.ResponseWriter, req *http.Request) (interface{}, error) {
var entMeta structs.EnterpriseMeta
if err := s.parseEntMetaNoWildcard(req, &entMeta); err != nil {
return nil, err
}
args := structs.IntentionQueryRequest{
Exact: &structs.IntentionQueryExact{},
}
if done := s.parse(resp, req, &args.Datacenter, &args.QueryOptions); done {
return nil, nil
}
q := req.URL.Query()
// Extract the source/destination
source, ok := q["source"]
if !ok || len(source) != 1 {
return nil, fmt.Errorf("required query parameter 'source' not set")
}
destination, ok := q["destination"]
if !ok || len(destination) != 1 {
return nil, fmt.Errorf("required query parameter 'destination' not set")
}
{
ap, ns, name, err := parseIntentionStringComponent(source[0], &entMeta)
if err != nil {
return nil, fmt.Errorf("source %q is invalid: %s", source[0], err)
}
args.Exact.SourcePartition = ap
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
4 years ago
args.Exact.SourceNS = ns
args.Exact.SourceName = name
}
{
ap, ns, name, err := parseIntentionStringComponent(destination[0], &entMeta)
if err != nil {
return nil, fmt.Errorf("destination %q is invalid: %s", destination[0], err)
}
args.Exact.DestinationPartition = ap
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
4 years ago
args.Exact.DestinationNS = ns
args.Exact.DestinationName = name
}
var reply structs.IndexedIntentions
if err := s.agent.RPC("Intention.Get", &args, &reply); err != nil {
// We have to check the string since the RPC sheds the error type
if err.Error() == consul.ErrIntentionNotFound.Error() {
resp.WriteHeader(http.StatusNotFound)
fmt.Fprint(resp, err.Error())
return nil, nil
}
// Not ideal, but there are a number of error scenarios that are not
// user error (400). We look for a specific case of invalid UUID
// to detect a parameter error and return a 400 response. The error
// is not a constant type or message, so we have to use strings.Contains
if strings.Contains(err.Error(), "UUID") {
return nil, BadRequestError{Reason: err.Error()}
}
return nil, err
}
// This shouldn't happen since the called API documents it shouldn't,
// but we check since the alternative if it happens is a panic.
if len(reply.Intentions) == 0 {
resp.WriteHeader(http.StatusNotFound)
return nil, nil
}
return reply.Intentions[0], nil
}
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
4 years ago
// PUT /v1/connect/intentions/exact
func (s *HTTPHandlers) IntentionPutExact(resp http.ResponseWriter, req *http.Request) (interface{}, error) {
var entMeta structs.EnterpriseMeta
if err := s.parseEntMetaNoWildcard(req, &entMeta); err != nil {
return nil, err
}
exact, err := parseIntentionQueryExact(req, &entMeta)
if err != nil {
return nil, err
}
args := structs.IntentionRequest{
Op: structs.IntentionOpUpsert,
}
s.parseDC(req, &args.Datacenter)
s.parseToken(req, &args.Token)
if err := decodeBody(req.Body, &args.Intention); err != nil {
return nil, BadRequestError{Reason: fmt.Sprintf("Request decode failed: %v", err)}
}
// Explicitly CLEAR the old legacy ID field
args.Intention.ID = ""
// Use the intention identity from the URL.
args.Intention.SourcePartition = exact.SourcePartition
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
4 years ago
args.Intention.SourceNS = exact.SourceNS
args.Intention.SourceName = exact.SourceName
args.Intention.DestinationPartition = exact.DestinationPartition
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
4 years ago
args.Intention.DestinationNS = exact.DestinationNS
args.Intention.DestinationName = exact.DestinationName
args.Intention.FillPartitionAndNamespace(&entMeta, false)
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
4 years ago
var ignored string
if err := s.agent.RPC("Intention.Apply", &args, &ignored); err != nil {
return nil, err
}
return true, nil
}
// DELETE /v1/connect/intentions/exact
func (s *HTTPHandlers) IntentionDeleteExact(resp http.ResponseWriter, req *http.Request) (interface{}, error) {
var entMeta structs.EnterpriseMeta
if err := s.parseEntMetaNoWildcard(req, &entMeta); err != nil {
return nil, err
}
exact, err := parseIntentionQueryExact(req, &entMeta)
if err != nil {
return nil, err
}
args := structs.IntentionRequest{
Op: structs.IntentionOpDelete,
Intention: &structs.Intention{
// NOTE: ID is explicitly empty here
SourcePartition: exact.SourcePartition,
SourceNS: exact.SourceNS,
SourceName: exact.SourceName,
DestinationPartition: exact.DestinationPartition,
DestinationNS: exact.DestinationNS,
DestinationName: exact.DestinationName,
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
4 years ago
},
}
s.parseDC(req, &args.Datacenter)
s.parseToken(req, &args.Token)
var ignored string
if err := s.agent.RPC("Intention.Apply", &args, &ignored); err != nil {
return nil, err
}
return true, nil
}
// intentionCreateResponse is the response structure for creating an intention.
type intentionCreateResponse struct{ ID string }
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
4 years ago
func parseIntentionQueryExact(req *http.Request, entMeta *structs.EnterpriseMeta) (*structs.IntentionQueryExact, error) {
q := req.URL.Query()
// Extract the source/destination
source, ok := q["source"]
if !ok || len(source) != 1 || source[0] == "" {
return nil, fmt.Errorf("required query parameter 'source' not set")
}
destination, ok := q["destination"]
if !ok || len(destination) != 1 || destination[0] == "" {
return nil, fmt.Errorf("required query parameter 'destination' not set")
}
var exact structs.IntentionQueryExact
{
ap, ns, name, err := parseIntentionStringComponent(source[0], entMeta)
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
4 years ago
if err != nil {
return nil, fmt.Errorf("source %q is invalid: %s", source[0], err)
}
exact.SourcePartition = ap
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
4 years ago
exact.SourceNS = ns
exact.SourceName = name
}
{
ap, ns, name, err := parseIntentionStringComponent(destination[0], entMeta)
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
4 years ago
if err != nil {
return nil, fmt.Errorf("destination %q is invalid: %s", destination[0], err)
}
exact.DestinationPartition = ap
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
4 years ago
exact.DestinationNS = ns
exact.DestinationName = name
}
return &exact, nil
}
func parseIntentionStringComponent(input string, entMeta *structs.EnterpriseMeta) (string, string, string, error) {
ss := strings.Split(input, "/")
switch len(ss) {
case 1: // Name only
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
4 years ago
ns := entMeta.NamespaceOrEmpty()
ap := entMeta.PartitionOrEmpty()
return ap, ns, ss[0], nil
case 2: // namespace/name
ap := entMeta.PartitionOrEmpty()
return ap, ss[0], ss[1], nil
case 3: // partition/namespace/name
return ss[0], ss[1], ss[2], nil
default:
return "", "", "", fmt.Errorf("input can contain at most two '/'")
}
}
// IntentionSpecific handles the endpoint for /v1/connect/intentions/:id.
// Deprecated: use IntentionExact.
func (s *HTTPHandlers) IntentionSpecific(resp http.ResponseWriter, req *http.Request) (interface{}, error) {
id := strings.TrimPrefix(req.URL.Path, "/v1/connect/intentions/")
switch req.Method {
case "GET":
return s.IntentionSpecificGet(id, resp, req)
case "PUT":
return s.IntentionSpecificUpdate(id, resp, req)
case "DELETE":
return s.IntentionSpecificDelete(id, resp, req)
default:
return nil, MethodNotAllowedError{req.Method, []string{"GET", "PUT", "DELETE"}}
}
}
// Deprecated: use IntentionGetExact.
func (s *HTTPHandlers) IntentionSpecificGet(id string, resp http.ResponseWriter, req *http.Request) (interface{}, error) {
// Method is tested in IntentionEndpoint
args := structs.IntentionQueryRequest{
IntentionID: id,
}
if done := s.parse(resp, req, &args.Datacenter, &args.QueryOptions); done {
return nil, nil
}
var reply structs.IndexedIntentions
if err := s.agent.RPC("Intention.Get", &args, &reply); err != nil {
// We have to check the string since the RPC sheds the error type
if err.Error() == consul.ErrIntentionNotFound.Error() {
resp.WriteHeader(http.StatusNotFound)
fmt.Fprint(resp, err.Error())
return nil, nil
}
// Not ideal, but there are a number of error scenarios that are not
// user error (400). We look for a specific case of invalid UUID
// to detect a parameter error and return a 400 response. The error
// is not a constant type or message, so we have to use strings.Contains
if strings.Contains(err.Error(), "UUID") {
return nil, BadRequestError{Reason: err.Error()}
}
return nil, err
}
// This shouldn't happen since the called API documents it shouldn't,
// but we check since the alternative if it happens is a panic.
if len(reply.Intentions) == 0 {
resp.WriteHeader(http.StatusNotFound)
return nil, nil
}
return reply.Intentions[0], nil
}
// Deprecated: use IntentionPutExact.
func (s *HTTPHandlers) IntentionSpecificUpdate(id string, resp http.ResponseWriter, req *http.Request) (interface{}, error) {
// Method is tested in IntentionEndpoint
var entMeta structs.EnterpriseMeta
if err := s.parseEntMetaNoWildcard(req, &entMeta); err != nil {
return nil, err
}
if entMeta.PartitionOrDefault() != structs.PartitionOrDefault("") {
return nil, BadRequestError{Reason: "Cannot use a partition with this endpoint"}
}
args := structs.IntentionRequest{
Op: structs.IntentionOpUpdate,
}
s.parseDC(req, &args.Datacenter)
s.parseToken(req, &args.Token)
if err := decodeBody(req.Body, &args.Intention); err != nil {
return nil, BadRequestError{Reason: fmt.Sprintf("Request decode failed: %v", err)}
}
if args.Intention.DestinationPartition != "" && args.Intention.DestinationPartition != "default" {
return nil, BadRequestError{Reason: "Cannot specify a destination partition with this endpoint"}
}
if args.Intention.SourcePartition != "" && args.Intention.SourcePartition != "default" {
return nil, BadRequestError{Reason: "Cannot specify a source partition with this endpoint"}
}
args.Intention.FillPartitionAndNamespace(&entMeta, false)
// Use the ID from the URL
args.Intention.ID = id
var reply string
if err := s.agent.RPC("Intention.Apply", &args, &reply); err != nil {
return nil, err
}
// Update uses the same create response
return intentionCreateResponse{reply}, nil
}
// Deprecated: use IntentionDeleteExact.
func (s *HTTPHandlers) IntentionSpecificDelete(id string, resp http.ResponseWriter, req *http.Request) (interface{}, error) {
// Method is tested in IntentionEndpoint
args := structs.IntentionRequest{
Op: structs.IntentionOpDelete,
Intention: &structs.Intention{ID: id},
}
s.parseDC(req, &args.Datacenter)
s.parseToken(req, &args.Token)
var reply string
if err := s.agent.RPC("Intention.Apply", &args, &reply); err != nil {
return nil, err
}
return true, nil
}