Skip to main content
Documentation for version v73

Communication Channels Workflow

General Structure​

When a campaign is launched, the campaign binary is executed. It:

  • processes the template and business logic (blocklists, limits, subscriptions, filters, etc.);
  • builds the message structure for delivery;
  • sends the prepared messages to RabbitMQ queues;
  • records errors in the campaign log.

Each message then goes to the corresponding processing service depending on the channel.

Email​

Several types of senders are supported for the email channel: AKMTA (the Altcraft sender) and external senders. The workflow may vary depending on the sender used.

Sending Process​

AKMTA​

The campaign sends messages to the following queues:

  • akmta_<sender_id> — standard queue;
  • akmta_prior_<sender_id> — priority queue.

Messages from the queues are processed by the akmtad service. Its behavior depends on the queue type:

  • For a non-priority queue, it writes messages to disk, periodically reads them, and sends them according to the sending strategy configured for the specific sender_id;
  • For a priority queue, it makes an attempt to send immediately; if it fails, it writes the message to disk.

The sending flow looks like this:

campaign → RMQ Queue (akmta / akmta_prior) → akmtad → SMTP

Other Senders​

External senders receive messages via an HTTP request from the campaign binary. If the platform cannot send a message, a campaign error is logged. Note that such an error is not considered a “Non-Delivery” and is not included in the non-delivery report.

Status Registration​

AKMTA​

Delivery statuses are registered immediately using:

  • the SMTP server response code;
  • bounce patterns (analysis of bounce templates).

Other Senders​

Tracking service of the platform handles event status.

Retry Logic​

AKMTA​

  • Retry rules are defined separately in the Admin Panel. Details are available in the administrator documentation;
  • Based on the SMTP code, hard bounce or soft bounce statuses are registered;
  • Retry logic differs for soft and hard bounce cases.

Other Senders​

External senders do not support retry attempts. If an HTTP request fails, the platform does nothing further.

SMS​

Sending Process​

The campaign sends messages to a single queue — sms_sender. The procintegras service then sends messages from the queue. Some gateways support batch sending.

Status Registration​

Status registration works in one of the following ways:

  • The gateway sends a message status via a webhook. The status is processed by procsmslisten;
  • The platform polls the gateway. The status is processed by procsmsev.

The lifetime of an SMS message is defined by the configuration parameter SMS_DEFAULT_TTL (default — 1 day). The platform registers a non-delivery if no status is received within this time.

Retry Logic​

SMS retries happen after 60, 120, 180, 300, and 600 seconds from the last failed attempt. The maximum number of attempts is 5.

Push​

Sending Process​

The campaign sends messages to a queue. The specific queue depends on the provider:

  • APNs — apns_push_sender
  • Firebase — firebase_push_sender
  • Huawei — huawei_push_sender
  • RuStore — rustore_push_sender
  • Browser push (Firefox, Safari) — web_push_sender
  • Other providers — custom_push_sender

The procintegras service then sends messages from the queue. Some gateways support batch sending.

Status Registration​

For browser push, the cookie_saver service handles status updates and sends events to push_events. Further processing occurs in procpush.

For mobile push, the logic is similar, but asynchronous sends are processed via push_events, while synchronous ones are immediately handled by procrpc.

The lifetime of a push message is defined by MTA_DEFAULT_TTL (default — 3 days). A non-delivery is registered if no status is received within this time.

Synchronous and Asynchronous Requests

Synchronous requests — the platform registers the result based on the response, meaning it waits for a reply.
Asynchronous requests — the platform registers the result when putting the message into the queue, meaning it does not wait for a reply.

Retry Logic​

The sender checks the gateway’s response and headers to apply retry rules. If the gateway explicitly specifies retry instructions (for example, “try again in 5 minutes”), retries follow them. If not, the platform uses its configured parameters:

  • PUSH_SENDER_RETRY_MAX_AMOUNT — maximum retry count (default — 3);
  • PUSH_SENDER_RETRY_INTERVAL — interval between retries (seconds) (default — 1800);
  • PUSH_SENDER_RETRY_SCAN_INTERVAL — scan interval for retry records (seconds) (default — 60).

Detailed information is available in the administrator documentation.

Messenger Channels​

Sending Process​

The campaign sends messages to a queue. RabbitMQ queues are processed depending on the channel:

  • Telegram Group, Telegram Bot — queue piper_messages, processed by procpiper;
  • Viber, WhatsApp, Notify — processed by procintegras.

Status Registration​

  • Telegram — status is determined based on the HTTP response from Telegram. Delivery is registered if no error occurs during the send request. A non-delivery is registered if an error is returned.
  • Viber, WhatsApp, Notify — statuses arrive via webhooks. If an error occurs during sending, a non-delivery event is registered. Statuses are handled by trklisten.

Retry Logic​

Messenger channels do not support retries at the moment.

Sending limits

Telegram imposes a limit of 28 requests per second per token.
For Notify, the default limit is 20 requests per second, configurable via VKNOTIFY_SEND_LIMIT_PER_SECOND, with a maximum of 50 requests per second.
For Viber, the default limit is 50 requests per second, configurable via VIBER_DEVINO_SEND_LIMIT_PER_SECOND.
For WhatsApp, the default limit is 20 requests per second (WHATSAPP_SEND_LIMIT_PER_SECOND); individual providers may apply their own request limits.

Last updated on