Planning

Notification channels

Set up Email, Slack, Microsoft Teams and webhook destinations for alerts and anomaly notifications, and test them.

Planned: email channels

Not yet available in Costfluent.

Planned: Slack channels

Not yet available in Costfluent.

Planned: Microsoft Teams channels

Not yet available in Costfluent.

Planned: webhook channels

Not yet available in Costfluent.

A notification channel is a destination Costfluent can send alerts to. Channels belong to the organization, so every workspace's alerts can use them. Open Settings, then Notifications.

Before you start

  • Adding, editing, testing and deleting channels needs the Owner, Integration owner or Editor organization role. A Viewer sees the channels without these actions.
  • Adding or editing a channel, Add to Slack included, and linking your Slack account need a recent sign-in.
  • A channel answers to the person who created it or last changed its destination, that is its email recipients or its Slack, Teams or webhook URL: through an API token, the token's creator; through a service token, the organization; through an app a member authorized, that member. Renaming, enabling or disabling a channel, or changing a webhook's events or signing secret, keeps whom it answers to, and so does installing the Slack app into the same channel again. It matters for two kinds of message, and Costfluent checks it each time one is sent, retries included: credit notices reach a channel only while that person can see the connection's credits, and the webhook events spend_band and subscription_status only while that person can manage billing, that is an organization Owner or, in an organization a partner manages, one of the partner's managers. A channel set up through a service token receives both, and a channel that answers to no one receives neither. Such a message is skipped and not tried again, so it shows as skipped under Last delivery.

Every channel type, Costfluent's Slack app included:

Included in every plan. Compare plans

Add a channel

The Notifications page listing three channels with their name, type and last delivery, Test, edit and delete actions on each, and Deliveries on the webhook

Notification channels for an organization with an email, a Slack and a webhook channel.
  1. Under Notifications, select Add channel.

  2. In Add notification channel, choose the Type, then enter a Name, such as Engineering alerts.

  3. Enter the destination for that type:

    • Email: the Email addresses, separated by commas. One email channel sends to at most 25 addresses. Addresses of people suspended in or removed from the organization are skipped when a message is sent; when every address on a channel is skipped, sending to it fails, and so does its test. Every email a channel sends carries an unsubscribe link: once a recipient uses it, no channel of the organization emails that address again, while other organizations' channels still can. If an email app cuts the link short, the page it opens says the link was not recognised and nothing changes; the link in any other email the organization's channels sent to that address still works. To addresses of people who are not active members of the organization, its channels send at most 1,000 emails in any 24 hours, and at most 250 to any one address; further emails are not sent, then or later.
    • Slack: under How Costfluent posts to Slack, keep Slack app to add Costfluent's Slack app, as the next section describes, or choose Incoming webhook and enter the Webhook URL of a Slack incoming webhook, an https address on hooks.slack.com; Get a webhook URL says how to make one. The Slack app takes no Name: it is named after the channel you pick in Slack.
    • Microsoft Teams: the Workflows webhook URL of a webhook from the Teams Workflows app; Get a webhook URL says how to make one. It must be an https address on a subdomain of environment.api.powerplatform.com. Office 365 connector webhooks are retired and are not accepted.
    • Webhook: the Webhook URL of your own endpoint. The request shows what Costfluent sends it. Under Events, tick the state changes the webhook should also receive; Choose state changes lists them.

    A webhook address is masked as you paste it; the eye beside it shows it, to check what you pasted.

    A Slack incoming webhook or Teams channel saved earlier on another address keeps sending and can still be renamed. A change to its configuration, through the Public API or a Terraform apply, needs an address the rules above accept.

  4. Select Add channel. A message confirms the channel was added; select Send test in it to send the channel a test message.

  5. For a webhook, Webhook channel added shows its Signing secret, which Costfluent generated for it. Copy it before you select Done: it is shown only this once. Webhook notifications describes how your endpoint checks a delivery with it.

The table lists each channel with its Name and Type. A deleted channel leaves the list. A Slack app channel removed in Slack stays, marked "Removed from Slack"; see How it behaves.

Get a webhook URL

A webhook URL is a credential: whoever has it can post to the channel. Costfluent stores it encrypted and never shows it again; a channel's read shows *** in its place.

  • Slack incoming webhook: at api.slack.com/apps, select Create New App, then From scratch, name it and choose your Slack workspace. Under Incoming Webhooks, turn on Activate Incoming Webhooks, select Add New Webhook to Workspace, choose the channel and select Allow. Copy the Webhook URL, which starts with https://hooks.slack.com/services/.
  • Microsoft Teams Workflows webhook: in Teams, open the channel's More options (⋯) menu and select Workflows. Choose the template Post to a channel when a webhook request is received, check the team and channel, and select Add workflow. Copy the URL Teams shows when the workflow is added.

Add the Slack app

Costfluent's Slack app posts into one Slack channel you choose, like an incoming webhook, and adds what a webhook cannot: buttons to acknowledge an anomaly or snooze a cost alert, and the /costfluent command to ask for a report's link.

The Add notification channel dialog with Slack as the type, Slack app chosen under How Costfluent posts to Slack, and the Add to Slack button

Add notification channel with Slack chosen: the Slack app or an incoming webhook, and Add to Slack.
  1. Under Notifications, select Add channel, choose Slack as the Type, and keep Slack app under How Costfluent posts to Slack. Where Costfluent's Slack app is not available, only an incoming webhook is offered.
  2. Select Add to Slack. Slack opens in the same window.
  3. In Slack, choose the Slack workspace and the one channel Costfluent posts to, and allow the app. It asks for five scopes: incoming-webhook to let you choose the channel, chat:write to post there, commands for /costfluent, and app_mentions:read and im:history to receive the messages that mention the app or are sent to it directly. Costfluent cannot read any other message in a channel.
  4. Slack sends you back to Notifications, where "Slack channel added" confirms it. When you cancel in Slack, or the install expired or was started in another browser, a message says so and no channel is added.

The channel is named after the Slack workspace and channel you chose, such as "Acme #finops"; edit it to rename it. Its Type is "Slack app", and below its name the table shows the channel and workspace, such as "#finops in Acme".

One install is one channel: to post to another channel, add the app again and choose that channel. Adding it again to a channel it already posts to keeps that channel, its name, and the alerts, workspaces and report schedules that use it, and gives Costfluent a new token for it.

A button or the /costfluent command acts in Costfluent as the member who used it, so each Slack user links their Slack account to their Costfluent account, once for each Slack app channel. Every organization role can link one.

  1. In Slack, select a button on a Costfluent message, or type /costfluent report and a report's name. The first time, Costfluent answers with a link in a message only you see. The link works for 10 minutes, and only once.
  2. Open the link, and sign in to Costfluent if it asks you to. Open it in the organization the Slack app belongs to: in another one, the link is refused, so switch organization and open it from Slack again.
  3. The page names the Slack user ID and the Slack workspace the link is for. To see your own member ID in Slack, open your profile, select the three dots, then Copy member ID. Check that they are yours and select Link account, or Not now to leave it. If you signed in more than 60 minutes ago, Confirm it's you opens first. Slack then asks you to sign in, which proves you are that Slack user: sign in as the Slack user the link names, in its workspace, and finish in the same browser. Your account is linked once Slack sends you back.
  4. "Slack account linked" confirms it. Go back to Slack and select the button or type the command again.

From then on, what you do through that Slack app runs as you, with your role in Costfluent at that moment. A link that expired or was used already shows "This Slack link cannot be used"; ask for a new one in Slack. If you sign in to Slack as anyone else, "This Slack link was made for another Slack user" shows and nothing is linked: select Sign in with Slack again and sign in as the Slack user the link names. If someone sent you the link, do not use it: ask them to open it themselves. When another member later links the same Slack user, which means signing in to Slack as that Slack user, the later link replaces the earlier one.

A new or reset password, a changed email address and Sign out everywhere else unlink your Slack account in every organization. When an Owner signs you out everywhere, or you leave or are removed from an organization, it is unlinked there. To use the Slack app again, link it from Slack as above.

Act from Slack

The Slack app posts anomalies and cost alerts with buttons, and everything else a Slack channel receives, such as budget thresholds and scheduled cost reports, as an incoming webhook posts it.

  • Acknowledge an anomaly: "Acknowledge" on a detected anomaly acknowledges it in Costfluent, as Insights does. An anomaly that is no longer open is left as it is, and Costfluent tells you its status.
  • Snooze a cost alert: "Snooze 1 day" and "Snooze 7 days" pause the alert for one day or seven, counted from when you select it. In Costfluent the alert shows "Paused until" and the time, and the first evaluation after that time resumes it on its own. Evaluations run every 15 minutes, so the alert can still show as paused for a short while after its snooze ends. Pausing the alert in Costfluent turns the snooze into a pause without an end, and resuming it ends the snooze. An alert paused in Costfluent without an end stays paused: a snooze leaves it, and Costfluent tells you so.
  • Ask for a report: /costfluent report <name> answers, only to you, with links to at most 10 reports whose title contains the name, from every workspace you can view in the Slack app's organization. The links open the reports in Costfluent, where you sign in; the answer holds no cost figures. Without report and a name, the command answers how to ask. In a channel the Slack app was not added to, it says that Costfluent is not installed there.

After you acknowledge or snooze, the message loses its buttons and gains a line saying who did it, and for a snooze, until when in UTC. Acknowledging and snoozing need the team role Editor with edit access to the workspace, as in Costfluent; with another role, Costfluent answers "Your role in Costfluent does not allow this." and changes nothing.

Terraform reads a snoozed alert as is_paused = true. When the alert is managed with Terraform and its configuration says is_paused = false, the next terraform apply resumes it and ends the snooze: the configuration wins.

Test a channel

Select Test on a channel to send a test message to it. "Test notification sent to …" confirms that Costfluent sent it; when sending fails, the message says why. Check that the message arrives before relying on the channel for alerts. Right after you add a channel, Send test in the message that confirms it does the same. Tests are rate limited per organization, counted across all its channels and members; once the limit is reached, a further test is refused for up to an hour.

A test is not recorded as a delivery. The Last delivery column shows each channel's newest real notification of the last 30 days: delivered and when, sending, which attempt failed and when the next one is due, failed after its last attempt, or skipped. A channel with nothing sent in 30 days reads "No deliveries in the last 30 days". Every member sees the column, counting the notifications for the workspaces they can see and those for no workspace.

Under a delivery that failed, is being retried or was skipped, a second line says why. For a webhook that is the HTTP status your endpoint answered, a redirect, which Costfluent does not follow, an address that is not public, no answer in time, or no connection to your endpoint. A skipped delivery names its reason, such as a plan that does not include the channel. Any other failure reads that Costfluent's team can see why; contact us if it keeps failing. The Owner, Integration owner and Editor roles see this line; a Viewer sees the status only.

See a webhook's deliveries

Select Deliveries on a webhook channel, enabled or disabled, to see what Costfluent sent it in the last 30 days, newest first.

The Deliveries panel of the Ops Webhook channel, describing the list, linking to how deliveries and resends work, and saying there were no deliveries in the last 30 days

Deliveries to a webhook that has had nothing sent in the last 30 days.
  1. Find a delivery by the ID your endpoint logged. Each delivery shows the value of its X-Costfluent-Delivery header, which is also delivery_id in the body, with its event type, when it was written and where it stands, and, as under Last delivery, why it failed or was skipped.
  2. Select Load more to read older deliveries.
  3. On a delivery that failed after its last attempt, select Resend to send it again.

A resent delivery starts the retry schedule again within a minute and keeps its ID, so an endpoint that already handled it can drop it. Each attempt carries a new timestamp and signature, as a retry does. Only a failed delivery can be resent, not one that is being sent, was delivered or was skipped. Every member sees the deliveries for the workspaces they can see and those for no workspace; every role but Viewer can resend. Resends are rate limited per organization; once the limit is reached, a further resend is refused for up to an hour.

Edit or delete a channel

  • Edit: the pencil opens Edit notification channel, where you can change the Name, and for a webhook its Events. For a webhook, Rotate signing secret replaces its signing secret with a new one Costfluent generates, after you confirm, and shows it once; the old one stops signing at once. Webhook notifications describes what your endpoint sees meanwhile. To change a destination, add a new channel and delete the old one.
  • Delete: the bin opens Delete channel, which lists what sends to the channel before you confirm: each cost alert, budget and report schedule by name with its workspace, and each workspace whose anomaly notifications or commitment expiry notices use it. Uses in workspaces you cannot see are counted, not named. Once it is deleted, you no longer receive notifications on that channel. Deleting it takes it out of every cost alert, budget, workspace's anomaly and commitment expiry notices and report schedule that named it, and a message says how many there were; one left with no channel sends nothing until you give it another. Deleting a Slack app channel also asks Slack to revoke Costfluent's token for it; if Slack does not answer, the channel is deleted anyway, and removing the app in Slack revokes the token.

How it behaves

  • Detected anomalies go to the channels their workspace chose under Anomaly notifications on the workspace's settings page, and to none when it chose none; see Anomaly notifications.
  • A commitment entering a workspace's lead time goes to the channels the workspace chose under Commitment expiry notices, once per term, and to none when it chose none; a channel two workspaces chose receives it once. The organization's owners also receive it by email. See Commitments.
  • Spend threshold, percentage change and tag coverage alerts go to the channels the alert names; see Alerts.
  • Budget thresholds go to the channels the budget names; see Budgets.
  • When a connection's data runs two days behind its sync schedule, every enabled channel is told, once more at seven days, and again only after the connection has caught up and fallen behind anew. While a new connection's first import is still running, it is measured from the day the connection was added, not from the older months the import has brought in so far. The notice names the connection, the last day its data covers and why the last sync failed, if it did. The person who added the connection, or the organization's owners when that person can no longer sign in, also receive it by email. A Kubernetes cluster's connection is told instead a day after its agent's last report, once more at three days, and again only after the agent has reported; see Kubernetes.
  • When a new connection's first import finishes, once every month its first collection asked for has arrived, each workspace the connection belongs to gets one notification in the bell, which opens the connection; a connection in no workspace notifies the whole organization. It names the months that arrived, how many cost records they hold and the last day the data covers, or says the connection works and found no cost to import. For AWS, when Costfluent read recent months from Cost Explorer, it adds that resource and tag detail follows with the export within 24 to 72 hours. The person who added the connection, or the organization's owners when that person can no longer sign in, also receive it by email. Of the channels, only a webhook subscribed to connection_imported receives it; see Webhooks.
  • Scheduled cost reports are delivered to the email, Slack and Microsoft Teams channels chosen on the schedule; a dashboard digest to its email channels only. See Share and schedule reports.
  • Each channel gets its own copy, usually within a minute. When a channel cannot be reached, Costfluent tries that channel again after 1, 5, 15, 60 and 240 minutes and stops after the sixth attempt. The channel's Last delivery shows each failed attempt until one succeeds. The other channels keep the copy they already have and are not sent another.
  • A channel that answers only after its own timeout can receive the same notification twice, once from the retry. Every attempt at one notification carries the same key: webhook requests carry it as delivery_id in the signed body and the X-Costfluent-Delivery header, and emails keep the same Message-Id, so your endpoint or mail system can drop the repeat.
  • A notification waiting for a channel that was deleted or disabled is not sent.
  • Removing Costfluent's app from a Slack workspace, or revoking its token in Slack, disables the Slack app channels it served there. They stay in the table marked "Removed from Slack" and send nothing, and the alerts, budgets, workspaces and report schedules that use them keep them. Such a channel offers only Add to Slack again and the bin. Add to Slack again, choosing the same Slack channel, enables the channel again for all of them; choosing another Slack channel adds a new channel instead. Deleting a removed channel does not ask Slack again: its token ended with the removal.
  • Slack messages, their buttons and Costfluent's answers in Slack are in English.