聊聊cortex的tenant-木庄网络博客

本文摘自网络，作者，侵删。

序

本文主要研究一下cortex的tenant

tenant

cortex/pkg/tenant/tenant.go

var (
    errTenantIDTooLong = errors.New("tenant ID is too long: max 150 characters")
)

type errTenantIDUnsupportedCharacter struct {
    pos      int
    tenantID string
}

func (e *errTenantIDUnsupportedCharacter) Error() string {
    return fmt.Sprintf(
        "tenant ID '%s' contains unsupported character '%c'",
        e.tenantID,
        e.tenantID[e.pos],
    )
}

const tenantIDsLabelSeparator = "|"

// NormalizeTenantIDs is creating a normalized form by sortiing and de-duplicating the list of tenantIDs
func NormalizeTenantIDs(tenantIDs []string) []string {
    sort.Strings(tenantIDs)

    count := len(tenantIDs)
    if count <= 1 {
        return tenantIDs
    }

    posOut := 1
    for posIn := 1; posIn < count; posIn++ {
        if tenantIDs[posIn] != tenantIDs[posIn-1] {
            tenantIDs[posOut] = tenantIDs[posIn]
            posOut++
        }
    }

    return tenantIDs[0:posOut]
}

// ValidTenantID
func ValidTenantID(s string) error {
    // check if it contains invalid runes
    for pos, r := range s {
        if !isSupported(r) {
            return &errTenantIDUnsupportedCharacter{
                tenantID: s,
                pos:      pos,
            }
        }
    }

    if len(s) > 150 {
        return errTenantIDTooLong
    }

    return nil
}

func JoinTenantIDs(tenantIDs []string) string {
    return strings.Join(tenantIDs, tenantIDsLabelSeparator)
}

// this checks if a rune is supported in tenant IDs (according to
// https://cortexmetrics.io/docs/guides/limitations/#tenant-id-naming)
func isSupported(c rune) bool {
    // characters
    if ('a' <= c && c <= 'z') || ('A' <= c && c <= 'Z') {
        return true
    }

    // digits
    if '0' <= c && c <= '9' {
        return true
    }

    // special
    return c == '!' ||
        c == '-' ||
        c == '_' ||
        c == '.' ||
        c == '*' ||
        c == '\'' ||
        c == '(' ||
        c == ')'
}

// TenantIDsFromOrgID extracts different tenants from an orgID string value
//
// ignore stutter warning
//nolint:golint
func TenantIDsFromOrgID(orgID string) ([]string, error) {
    return TenantIDs(user.InjectOrgID(context.TODO(), orgID))
}

tenant提供了NormalizeTenantIDs、ValidTenantID、JoinTenantIDs方法；NormalizeTenantIDs用于去重和排序tenantIDs；ValidTenantID会通过isSupported进行校验，并校验长度；JoinTenantIDs使用|来连接TenantID

Resolver

cortex/pkg/tenant/resolver.go

type Resolver interface {
    // TenantID returns exactly a single tenant ID from the context. It should be
    // used when a certain endpoint should only support exactly a single
    // tenant ID. It returns an error user.ErrNoOrgID if there is no tenant ID
    // supplied or user.ErrTooManyOrgIDs if there are multiple tenant IDs present.
    TenantID(context.Context) (string, error)

    // TenantIDs returns all tenant IDs from the context. It should return
    // normalized list of ordered and distinct tenant IDs (as produced by
    // NormalizeTenantIDs).
    TenantIDs(context.Context) ([]string, error)
}

Resolver接口定义了TenantID、TenantIDs方法

SingleResolver

cortex/pkg/tenant/resolver.go

type SingleResolver struct {
}

func (t *SingleResolver) TenantID(ctx context.Context) (string, error) {
    //lint:ignore faillint wrapper around upstream method
    return user.ExtractOrgID(ctx)
}

func (t *SingleResolver) TenantIDs(ctx context.Context) ([]string, error) {
    //lint:ignore faillint wrapper around upstream method
    orgID, err := user.ExtractOrgID(ctx)
    if err != nil {
        return nil, err
    }
    return []string{orgID}, err
}

SingleResolver实现了Resolver接口，其TenantID使用user.ExtractOrgID(ctx)提取TenantID；TenantIDs方法则返回的是orgID数组

MultiResolver

cortex/pkg/tenant/resolver.go

type MultiResolver struct {
}

// NewMultiResolver creates a tenant resolver, which allows request to have
// multiple tenant ids submitted separated by a '|' character. This enforces
// further limits on the character set allowed within tenants as detailed here:
// https://cortexmetrics.io/docs/guides/limitations/#tenant-id-naming)
func NewMultiResolver() *MultiResolver {
    return &MultiResolver{}
}

func (t *MultiResolver) TenantID(ctx context.Context) (string, error) {
    orgIDs, err := t.TenantIDs(ctx)
    if err != nil {
        return "", err
    }

    if len(orgIDs) > 1 {
        return "", user.ErrTooManyOrgIDs
    }

    return orgIDs[0], nil
}

func (t *MultiResolver) TenantIDs(ctx context.Context) ([]string, error) {
    //lint:ignore faillint wrapper around upstream method
    orgID, err := user.ExtractOrgID(ctx)
    if err != nil {
        return nil, err
    }

    orgIDs := strings.Split(orgID, tenantIDsLabelSeparator)
    for _, orgID := range orgIDs {
        if err := ValidTenantID(orgID); err != nil {
            return nil, err
        }
    }

    return NormalizeTenantIDs(orgIDs), nil
}

MultiResolver实现了Resolver接口，其TenantID方法使用TenantIDs提取orgIDs，并校验其长度；TenantIDs方法使用user.ExtractOrgID(ctx)提取orgID，然后使用tenantIDsLabelSeparator分割提取为orgIDs，最后通过NormalizeTenantIDs返回

小结

cortex的tenant提供了NormalizeTenantIDs、ValidTenantID、JoinTenantIDs方法；Resolver接口定义了TenantID、TenantIDs方法；SingleResolver及MultiResolver都实现了Resolver接口，基本都是通过user.ExtractOrgID(ctx)提取tenantID。