import dayjs from 'dayjs' import { AlertTriangle, CheckCircle2, ChevronRight, Loader2 } from 'lucide-react' import Link from 'next/link' import { useParams } from 'common' import { InlineLink } from 'components/ui/InlineLink' import { SingleStat } from 'components/ui/SingleStat' import { useBranchesQuery } from 'data/branches/branches-query' import { useEdgeFunctionServiceStatusQuery } from 'data/service-status/edge-functions-status-query' import { ProjectServiceStatus, useProjectServiceStatusQuery, } from 'data/service-status/service-status-query' import { useIsFeatureEnabled } from 'hooks/misc/useIsFeatureEnabled' import { useSelectedProjectQuery } from 'hooks/misc/useSelectedProject' import { DOCS_URL } from 'lib/constants' import { InfoIcon, PopoverContent_Shadcn_, PopoverTrigger_Shadcn_, Popover_Shadcn_, cn } from 'ui' const SERVICE_STATUS_THRESHOLD = 5 // minutes /** * [Joshen] JFYI before we go live with this, we need to revisit the migrations section * as I don't think it should live in the ServiceStatus component since its not indicative * of a project's "service". ServiceStatus's intention is to be an ongoing health/status check. * * For context, migrations are meant to be indicative for only when creating branches or projects * with an initial SQL, so "healthy" migrations just means that migrations have all been successfully * ran. So it might be a matter of decoupling "ready" state vs "health checks" * [Edit] Now that migrations are only showing up if the project is a branch, i think its okay for now * * [Joshen] Another issue that requires investigation before we go live with the changes: * We've removed the isProjectNew check in this component which we had that logic cause new * projects would show unhealthy as the services are still starting up - but it causes a * perceived negative impression as new projects were showing unhealthy, hence the 5 minute * threshold check (we’d show “Coming up” instead of “unhealthy” if the project is within 5 * minutes of when it was created). Might be related to decoupling "ready" state vs "health checks" */ const StatusMessage = ({ status, isLoading, isHealthy, isProjectNew, }: { isLoading: boolean isHealthy: boolean isProjectNew: boolean status?: ProjectServiceStatus }) => { if (isHealthy || status === 'ACTIVE_HEALTHY') return 'Healthy' if (isLoading) return 'Checking status' if (status === 'UNHEALTHY') return 'Unhealthy' if (isProjectNew || status === 'COMING_UP') return 'Coming up...' if (status) return status return 'Unable to connect' } const iconProps = { size: 18, strokeWidth: 1.5, } const LoaderIcon = () => const AlertIcon = () => const CheckIcon = () => const StatusIcon = ({ isLoading, isHealthy, isProjectNew, projectStatus, }: { isLoading: boolean isHealthy: boolean isProjectNew: boolean projectStatus?: ProjectServiceStatus }) => { if (isHealthy || projectStatus === 'ACTIVE_HEALTHY') return if (isLoading || isProjectNew || projectStatus === 'COMING_UP') return return } export const ServiceStatus = () => { const { ref } = useParams() const { data: project } = useSelectedProjectQuery() const { projectAuthAll: authEnabled, projectEdgeFunctionAll: edgeFunctionsEnabled, realtimeAll: realtimeEnabled, projectStorageAll: storageEnabled, } = useIsFeatureEnabled([ 'project_auth:all', 'project_edge_function:all', 'realtime:all', 'project_storage:all', ]) const isBranch = project?.parentRef !== project?.ref // Get branches data when on a branch const { data: branches, isPending: isBranchesLoading } = useBranchesQuery( { projectRef: isBranch ? project?.parentRef : undefined }, { enabled: isBranch, } ) const currentBranch = isBranch ? branches?.find((branch) => branch.project_ref === ref) : undefined // [Joshen] Need pooler service check eventually const { data: status, isPending: isLoading } = useProjectServiceStatusQuery( { projectRef: ref }, { refetchInterval: (query) => query.state.data?.some((service) => !service.healthy) ? 5000 : false, } ) const { data: edgeFunctionsStatus } = useEdgeFunctionServiceStatusQuery( { projectRef: ref }, { refetchInterval: (query) => (!query.state.data?.healthy ? 5000 : false) } ) const authStatus = status?.find((service) => service.name === 'auth') const restStatus = status?.find((service) => service.name === 'rest') const realtimeStatus = status?.find((service) => service.name === 'realtime') const storageStatus = status?.find((service) => service.name === 'storage') const dbStatus = status?.find((service) => service.name === 'db') const isMigrationLoading = project?.status === 'COMING_UP' || (isBranch && (isBranchesLoading || currentBranch?.status === 'CREATING_PROJECT' || currentBranch?.status === 'RUNNING_MIGRATIONS')) // [Joshen] Need individual troubleshooting docs for each service eventually for users to self serve const services: { name: string error?: string docsUrl?: string isLoading: boolean isHealthy: boolean status: ProjectServiceStatus logsUrl: string }[] = [ { name: 'Database', error: undefined, docsUrl: undefined, isLoading: isLoading, isHealthy: !!dbStatus?.healthy, status: dbStatus?.status ?? 'UNHEALTHY', logsUrl: '/logs/postgres-logs', }, { name: 'PostgREST', error: restStatus?.error, docsUrl: undefined, isLoading, isHealthy: !!restStatus?.healthy, status: restStatus?.status ?? 'UNHEALTHY', logsUrl: '/logs/postgrest-logs', }, ...(authEnabled ? [ { name: 'Auth', error: authStatus?.error, docsUrl: undefined, isLoading, isHealthy: !!authStatus?.healthy, status: authStatus?.status ?? 'UNHEALTHY', logsUrl: '/logs/auth-logs', }, ] : []), ...(realtimeEnabled ? [ { name: 'Realtime', error: realtimeStatus?.error, docsUrl: undefined, isLoading, isHealthy: !!realtimeStatus?.healthy, status: realtimeStatus?.status ?? 'UNHEALTHY', logsUrl: '/logs/realtime-logs', }, ] : []), ...(storageEnabled ? [ { name: 'Storage', error: storageStatus?.error, docsUrl: undefined, isLoading, isHealthy: !!storageStatus?.healthy, status: storageStatus?.status ?? 'UNHEALTHY', logsUrl: '/logs/storage-logs', }, ] : []), ...(edgeFunctionsEnabled ? [ { name: 'Edge Functions', error: undefined, docsUrl: `${DOCS_URL}/guides/functions/troubleshooting`, isLoading, isHealthy: !!edgeFunctionsStatus?.healthy, status: edgeFunctionsStatus?.healthy ? 'ACTIVE_HEALTHY' : isLoading ? 'COMING_UP' : ('UNHEALTHY' as ProjectServiceStatus), logsUrl: '/logs/edge-functions-logs', }, ] : []), ...(isBranch ? [ { name: 'Migrations', error: undefined, docsUrl: undefined, isLoading: isBranchesLoading, isHealthy: isBranch ? currentBranch?.status === 'FUNCTIONS_DEPLOYED' : !isMigrationLoading, status: (isBranch ? currentBranch?.status === 'FUNCTIONS_DEPLOYED' ? 'ACTIVE_HEALTHY' : currentBranch?.status === 'FUNCTIONS_FAILED' || currentBranch?.status === 'MIGRATIONS_FAILED' ? 'UNHEALTHY' : 'COMING_UP' : isMigrationLoading ? 'COMING_UP' : 'ACTIVE_HEALTHY') as ProjectServiceStatus, logsUrl: isBranch ? '/branches' : '/logs/database-logs', }, ] : []), ] const isLoadingChecks = services.some((service) => service.isLoading) const allServicesOperational = services.every((service) => service.isHealthy) // Check if project or branch is in a startup state const isProjectNew = dayjs.utc().diff(dayjs.utc(project?.inserted_at), 'minute') < SERVICE_STATUS_THRESHOLD || project?.status === 'COMING_UP' || (isBranch && (currentBranch?.status === 'CREATING_PROJECT' || currentBranch?.status === 'RUNNING_MIGRATIONS' || isMigrationLoading)) const anyUnhealthy = services.some( (service) => !service.isHealthy && service.status !== 'COMING_UP' ) const anyComingUp = services.some((service) => service.status === 'COMING_UP') // Spinner only while the overall project is in COMING_UP; otherwise show 6-dot grid const showSpinnerIcon = project?.status === 'COMING_UP' const getOverallStatusLabel = (): string => { if (isLoadingChecks) return 'Checking...' if (anyComingUp) return 'Coming up...' if (anyUnhealthy) return 'Unhealthy' return 'Healthy' } const overallStatusLabel = getOverallStatusLabel() return ( ) : (
{services.map((service, index) => (
))}
) } label={Status} value={{overallStatusLabel}} /> {services.map((service) => (

{service.name}

View logs
))} {!allServicesOperational && (

{isProjectNew ? 'New' : 'Recently restored'} projects can take up to{' '} {SERVICE_STATUS_THRESHOLD} minutes to become fully operational.

If services stay unhealthy, refer to our{' '} docs {' '} for more information.

)}
) }