towerops/lib/towerops_web/live/agent_live/index.ex
Graham McIntie 1590f78bdc fix: input validation, SSRF, API hardening, and cookie security
- LIKE wildcard injection: sanitize %, _ in search queries (devices, sites, gaiia)
- Jason.decode! → Jason.decode with error handling for untrusted input
- inspect() leak: replace with generic error messages, log details server-side
- SSRF protection: URL validator blocks private IPs, localhost, non-HTTP schemes
- SSRF validation added to HTTP monitoring executor and integration credentials
- GraphQL complexity limits: always applied, not just in prod
- GraphQL introspection: also check GET query params, not just body
- Stripe webhook: explicit nil/empty checks for signature and body
- Cookie security: secure flag for session (prod), http_only+secure for remember_me
- Honeybadger API key: read from env var with fallback
- String.to_integer → Integer.parse with fallback for URL params
- String.to_atom → whitelist map for HTTP methods
- Gaiia webhook: remove secret_len and expected signature from log
- Admin API: add rate limiting pipeline
- to_atom_keys: per-key fallback instead of all-or-nothing rescue
2026-03-14 14:48:59 -05:00

491 lines
17 KiB
Elixir

defmodule ToweropsWeb.AgentLive.Index do
@moduledoc false
use ToweropsWeb, :live_view
use Gettext, backend: ToweropsWeb.Gettext
import ToweropsWeb.AgentLive.Helpers
import ToweropsWeb.GettextHelpers
alias Towerops.Accounts.Scope
alias Towerops.Agents
alias Towerops.Agents.Stats
alias Towerops.Settings
@impl true
def mount(_params, _session, socket) do
organization = socket.assigns.current_scope.organization
current_scope = socket.assigns.current_scope
is_superuser = Scope.superuser?(current_scope)
agent_tokens = Agents.list_organization_agent_tokens(organization.id)
# Subscribe to agent health updates for real-time status changes
timer_ref =
if connected?(socket) do
_ = Phoenix.PubSub.subscribe(Towerops.PubSub, "agents:health")
{:ok, ref} = :timer.send_interval(1000, :tick)
ref
end
# If superadmin (including when impersonating), also load cloud pollers and global default
{cloud_pollers, global_default_cloud_poller_id} =
if is_superuser do
{Agents.list_cloud_pollers(), Settings.get_global_default_cloud_poller()}
else
{[], nil}
end
# device counts for each agent (both direct and total with inheritance)
equipment_counts = calculate_device_counts(agent_tokens)
# device counts for cloud pollers (if superuser)
cloud_poller_counts =
if is_superuser do
calculate_device_counts(cloud_pollers)
else
%{}
end
# Get organization-wide agent health statistics
agent_health_stats = Stats.get_organization_agent_health(organization.id)
assignment_breakdown = Stats.get_device_assignment_breakdown(organization.id)
offline_agents = Stats.get_offline_agents(organization.id)
# Get agent image URL from config or use default
agent_image = Application.get_env(:towerops, :agent_docker_image, "towerops/agent:latest")
{:ok,
socket
|> assign(:page_title, t("Remote Agents"))
|> assign(:timezone, socket.assigns.current_scope.timezone)
|> assign(:is_superuser, is_superuser)
|> stream(:agent_tokens, agent_tokens)
|> assign(:has_agents, agent_tokens != [])
|> stream(:cloud_pollers, cloud_pollers)
|> assign(:cloud_pollers_list, cloud_pollers)
|> assign(:has_cloud_pollers, cloud_pollers != [])
|> assign(:global_default_cloud_poller_id, global_default_cloud_poller_id)
|> assign(:selected_global_default, global_default_cloud_poller_id || "")
|> assign(:device_counts, equipment_counts)
|> assign(:cloud_poller_counts, cloud_poller_counts)
|> assign(:agent_health_stats, agent_health_stats)
|> assign(:assignment_breakdown, assignment_breakdown)
|> assign(:offline_agents, offline_agents)
|> assign(:agent_image, agent_image)
|> assign(:new_token, nil)
|> assign(:show_token_modal, false)
|> assign(:now, DateTime.utc_now())
|> assign(:timer_ref, timer_ref)
|> assign(:agent_form, to_form(%{"name" => "", "is_cloud_poller" => false}))}
end
@impl true
def handle_params(params, _url, socket) do
# Read modal state from URL
modal = params["modal"]
socket =
socket
|> assign(:show_token_modal, modal == "setup")
|> apply_action(socket.assigns.live_action, params)
{:noreply, socket}
end
@impl true
def handle_event("create_agent", params, socket) do
{name, is_cloud_poller} = parse_agent_params(params)
with :ok <- validate_cloud_poller_permission(socket.assigns.current_scope, is_cloud_poller),
{:ok, agent_token, token} <- create_agent(socket.assigns.current_scope.organization, name, is_cloud_poller) do
handle_agent_creation_success(socket, agent_token, token, is_cloud_poller)
else
{:error, :unauthorized} ->
{:noreply, put_flash(socket, :error, t_equipment("Only superadmins can create cloud pollers"))}
{:error, _changeset} ->
error_message =
if is_cloud_poller,
do: t_equipment("Failed to create cloud poller"),
else: t_equipment("Failed to create agent")
{:noreply, put_flash(socket, :error, error_message)}
end
end
@impl true
def handle_event("close_token_modal", _params, socket) do
{:noreply,
socket
|> assign(:new_token, nil)
|> push_patch(to: ~p"/agents")}
end
@impl true
def handle_event("show_setup", %{"id" => id}, socket) do
agent_token = Agents.get_agent_token!(id)
{:noreply,
socket
|> assign(:new_token, %{agent_token: agent_token, token: agent_token.token})
|> push_patch(to: ~p"/agents?modal=setup")}
end
@impl true
def handle_event("regenerate_token", %{"id" => id}, socket) do
agent_token = Agents.get_agent_token!(id)
organization = socket.assigns.current_scope.organization
# Revoke the old token and create a new one
case Agents.revoke_agent_token(id) do
{:ok, _} ->
case Agents.create_agent_token(organization.id, agent_token.name) do
{:ok, new_agent_token, token} ->
agent_tokens = Agents.list_organization_agent_tokens(organization.id)
{:noreply,
socket
|> stream(:agent_tokens, agent_tokens, reset: true)
|> assign(:has_agents, agent_tokens != [])
|> assign(:new_token, %{agent_token: new_agent_token, token: token})
|> assign(:show_token_modal, true)
|> put_flash(:info, t_equipment("New token generated successfully"))}
{:error, _changeset} ->
{:noreply, put_flash(socket, :error, t_equipment("Failed to generate new token"))}
end
{:error, _} ->
{:noreply, put_flash(socket, :error, t_equipment("Failed to revoke old token"))}
end
end
@impl true
def handle_event("delete_agent", %{"id" => id}, socket) do
organization = socket.assigns.current_scope.organization
# Verify the agent token belongs to the current org before deleting
agent_token = Agents.get_agent_token!(id)
if agent_token.organization_id != organization.id do
{:noreply, put_flash(socket, :error, t_equipment("Agent not found"))}
else
case Agents.delete_agent_token(id) do
{:ok, _} ->
# If the deleted agent was the global default, clear it
if Scope.superuser?(socket.assigns.current_scope) &&
socket.assigns.global_default_cloud_poller_id == id do
Settings.set_global_default_cloud_poller(nil)
end
agent_tokens = Agents.list_organization_agent_tokens(organization.id)
# Refresh cloud pollers list if superadmin (in case a cloud poller was deleted)
cloud_pollers = load_cloud_pollers_if_superuser(socket.assigns.current_scope)
# Reload global default cloud poller ID
global_default_cloud_poller_id =
if Scope.superuser?(socket.assigns.current_scope) do
Settings.get_global_default_cloud_poller()
else
socket.assigns.global_default_cloud_poller_id
end
# Recalculate device counts after deletion
equipment_counts =
Map.new(agent_tokens, fn token ->
direct = Agents.count_assigned_devices(token.id)
total = length(Agents.list_agent_polling_targets(token.id))
{token.id, %{direct: direct, total: total}}
end)
cloud_poller_counts = calculate_device_counts(cloud_pollers)
# Refresh health statistics
agent_health_stats = Stats.get_organization_agent_health(organization.id)
assignment_breakdown = Stats.get_device_assignment_breakdown(organization.id)
offline_agents = Stats.get_offline_agents(organization.id)
{:noreply,
socket
|> stream(:agent_tokens, agent_tokens, reset: true)
|> assign(:has_agents, agent_tokens != [])
|> stream(:cloud_pollers, cloud_pollers, reset: true)
|> assign(:cloud_pollers_list, cloud_pollers)
|> assign(:has_cloud_pollers, cloud_pollers != [])
|> assign(:global_default_cloud_poller_id, global_default_cloud_poller_id)
|> assign(:selected_global_default, global_default_cloud_poller_id || "")
|> assign(:device_counts, equipment_counts)
|> assign(:cloud_poller_counts, cloud_poller_counts)
|> assign(:agent_health_stats, agent_health_stats)
|> assign(:assignment_breakdown, assignment_breakdown)
|> assign(:offline_agents, offline_agents)
|> put_flash(
:info,
t_equipment("Agent deleted successfully. Devices now fall back to site/org defaults or cloud polling.")
)}
{:error, _} ->
{:noreply, put_flash(socket, :error, t_equipment("Failed to delete agent"))}
end
end
end
@impl true
def handle_event("update_selected_global_default", %{"agent_token_id" => agent_token_id}, socket) do
# Just update the selected value in the dropdown, don't save yet
{:noreply, assign(socket, :selected_global_default, agent_token_id)}
end
@impl true
def handle_event("save_global_default", _params, socket) do
current_scope = socket.assigns.current_scope
if Scope.superuser?(current_scope) do
agent_token_id = socket.assigns.selected_global_default
# Handle empty string as nil
agent_token_id = if agent_token_id == "", do: nil, else: agent_token_id
# Validate that the agent exists if not nil
case validate_and_save_global_default(agent_token_id) do
{:ok, validated_id} ->
handle_global_default_success(socket, validated_id)
{:error, :agent_not_found} ->
{:noreply,
put_flash(socket, :error, t_equipment("Selected agent no longer exists. Please choose another agent."))}
{:error, _} ->
{:noreply, put_flash(socket, :error, t_equipment("Failed to update global default cloud poller"))}
end
else
{:noreply, put_flash(socket, :error, t_equipment("Only superadmins can set the global default cloud poller"))}
end
end
defp apply_action(socket, :index, _params) do
socket
end
defp parse_agent_params(params) do
case params do
%{"agent_form" => %{"name" => n, "is_cloud_poller" => cp}} ->
{n, cp in ["true", "on", true]}
%{"agent_form" => %{"name" => n}} ->
{n, false}
%{"name" => n, "is_cloud_poller" => cp} ->
{n, cp in ["true", "on", true]}
%{"name" => n} ->
{n, false}
end
end
defp validate_cloud_poller_permission(scope, is_cloud_poller) do
if is_cloud_poller && !Scope.superuser?(scope) do
{:error, :unauthorized}
else
:ok
end
end
defp create_agent(organization, name, is_cloud_poller) do
if is_cloud_poller do
Agents.create_cloud_poller(name)
else
Agents.create_agent_token(organization.id, name)
end
end
defp handle_agent_creation_success(socket, agent_token, token, is_cloud_poller) do
organization = socket.assigns.current_scope.organization
agent_tokens = Agents.list_organization_agent_tokens(organization.id)
cloud_pollers = load_cloud_pollers_if_superuser(socket.assigns.current_scope)
equipment_counts = calculate_device_counts(agent_tokens)
cloud_poller_counts = calculate_device_counts(cloud_pollers)
agent_health_stats = Stats.get_organization_agent_health(organization.id)
assignment_breakdown = Stats.get_device_assignment_breakdown(organization.id)
offline_agents = Stats.get_offline_agents(organization.id)
success_message = get_success_message(is_cloud_poller)
{:noreply,
socket
|> stream(:agent_tokens, agent_tokens, reset: true)
|> assign(:has_agents, agent_tokens != [])
|> stream(:cloud_pollers, cloud_pollers, reset: true)
|> assign(:cloud_pollers_list, cloud_pollers)
|> assign(:has_cloud_pollers, cloud_pollers != [])
|> assign(:device_counts, equipment_counts)
|> assign(:cloud_poller_counts, cloud_poller_counts)
|> assign(:agent_health_stats, agent_health_stats)
|> assign(:assignment_breakdown, assignment_breakdown)
|> assign(:offline_agents, offline_agents)
|> assign(:new_token, %{agent_token: agent_token, token: token})
|> assign(:agent_form, to_form(%{"name" => "", "is_cloud_poller" => false}))
|> put_flash(:info, success_message)
|> push_patch(to: ~p"/agents?modal=setup")}
end
defp load_cloud_pollers_if_superuser(scope) do
if Scope.superuser?(scope), do: Agents.list_cloud_pollers(), else: []
end
defp calculate_device_counts(agent_tokens) do
Map.new(agent_tokens, fn t ->
direct = Agents.count_assigned_devices(t.id)
total = length(Agents.list_agent_polling_targets(t.id))
{t.id, %{direct: direct, total: total}}
end)
end
defp get_success_message(is_cloud_poller) do
if is_cloud_poller,
do: t_equipment("Cloud poller created successfully"),
else: t_equipment("Agent created successfully")
end
defp validate_and_save_global_default(nil) do
# Clearing the default is always valid
case Settings.set_global_default_cloud_poller(nil) do
{:ok, _} -> {:ok, nil}
error -> error
end
end
defp validate_and_save_global_default(agent_token_id) do
# Verify the agent exists before saving
%Agents.AgentToken{} = Agents.get_agent_token!(agent_token_id)
case Settings.set_global_default_cloud_poller(agent_token_id) do
{:ok, _} -> {:ok, agent_token_id}
error -> error
end
rescue
Ecto.NoResultsError ->
{:error, :agent_not_found}
end
defp handle_global_default_success(socket, agent_token_id) do
message =
if agent_token_id,
do: t_equipment("Global default cloud poller set successfully"),
else: t_equipment("Global default cloud poller cleared")
{:noreply,
socket
|> assign(:global_default_cloud_poller_id, agent_token_id)
|> assign(:selected_global_default, agent_token_id || "")
|> put_flash(:info, message)}
end
@impl true
def handle_info(:tick, socket) do
{:noreply, assign(socket, :now, DateTime.utc_now())}
end
# Handle real-time agent status updates
# Only process events for this organization's agents, or cloud pollers if superuser
@impl true
def handle_info({:agent_connected, agent_token_id, organization_id}, socket) do
{:noreply, maybe_update_agent(socket, agent_token_id, organization_id)}
end
@impl true
def handle_info({:agent_disconnected, agent_token_id, organization_id}, socket) do
{:noreply, maybe_update_agent(socket, agent_token_id, organization_id)}
end
@impl true
def handle_info({:agent_heartbeat, agent_token_id, organization_id}, socket) do
{:noreply, maybe_update_agent(socket, agent_token_id, organization_id)}
end
@impl true
def handle_info({:agents_stale, _stale_agents}, socket) do
{:noreply, reload_agent_data(socket)}
end
defp reload_agent_data(socket) do
organization = socket.assigns.current_scope.organization
agent_tokens = Agents.list_organization_agent_tokens(organization.id)
cloud_pollers = load_cloud_pollers_if_superuser(socket.assigns.current_scope)
equipment_counts = calculate_device_counts(agent_tokens)
cloud_poller_counts = calculate_device_counts(cloud_pollers)
agent_health_stats = Stats.get_organization_agent_health(organization.id)
assignment_breakdown = Stats.get_device_assignment_breakdown(organization.id)
offline_agents = Stats.get_offline_agents(organization.id)
socket
|> stream(:agent_tokens, agent_tokens, reset: true)
|> assign(:has_agents, agent_tokens != [])
|> stream(:cloud_pollers, cloud_pollers, reset: true)
|> assign(:cloud_pollers_list, cloud_pollers)
|> assign(:has_cloud_pollers, cloud_pollers != [])
|> assign(:device_counts, equipment_counts)
|> assign(:cloud_poller_counts, cloud_poller_counts)
|> assign(:agent_health_stats, agent_health_stats)
|> assign(:assignment_breakdown, assignment_breakdown)
|> assign(:offline_agents, offline_agents)
end
defp maybe_update_agent(socket, agent_token_id, organization_id) do
my_org_id = socket.assigns.current_scope.organization.id
agent_token = Agents.get_agent_token!(agent_token_id)
cond do
agent_token.is_cloud_poller ->
# Cloud pollers only visible to superusers
if socket.assigns.is_superuser do
socket
|> stream_insert(:cloud_pollers, agent_token)
|> update_health_stats()
else
socket
end
organization_id == my_org_id ->
# Regular agent belonging to this organization
socket
|> stream_insert(:agent_tokens, agent_token)
|> update_health_stats()
true ->
# Agent from another organization, ignore
socket
end
rescue
Ecto.NoResultsError ->
# Agent was deleted, just refresh stats
update_health_stats(socket)
end
defp update_health_stats(socket) do
organization = socket.assigns.current_scope.organization
agent_health_stats = Stats.get_organization_agent_health(organization.id)
offline_agents = Stats.get_offline_agents(organization.id)
socket
|> assign(:agent_health_stats, agent_health_stats)
|> assign(:offline_agents, offline_agents)
end
@impl true
def terminate(_reason, socket) do
if timer_ref = socket.assigns[:timer_ref] do
:timer.cancel(timer_ref)
end
:ok
end
end