Sentry Alerts for Small Teams: Setup and Best Practices

To make Sentry alerts useful in a small team, page a person only for high-priority issues in production, send everything else to a channel or a ticket, and throttle repeats. In Sentry's current Alert builder that means a few rules, each with an environment filter, a priority or frequency filter, and an action interval set deliberately.
Key takeaways
- Sentry now splits the job in two: Monitors decide when a problem becomes an issue, and Alerts decide what to do about it. Most noise comes from alerts that act on every issue in every environment.
- Use three tiers: a page for production issues that hurt users now, a team channel for new and regressed issues, and a ticket for the rest.
- The action interval defaults to notifying on every trigger. Set it on every rule.
- Frequency filters (events, users affected, percentage of sessions) only work with the "new issue" and "event captured" triggers. Plan rules around that.
- Uptime monitors create issues but do not notify on their own. You need an alert that matches outage issues.
- Use Send Test Notification before you trust a rule, then review which alerts nobody acted on.
This guide is for small engineering teams, two to ten people, who use Sentry and either ignore most of its notifications or worry they will miss the one that matters. Every Sentry setting below is taken from Sentry's documentation as it reads on 13 November 2026; the rule set and thresholds are our recommendations.
What are Sentry alerts and monitors now?
Sentry has split its alerting into two parts. In its own words, it separates what you detect from what you do about it: Monitors watch your application and decide when a problem becomes an issue, and Alerts take action when issues match rules you define. An alert can send notifications, create tickets, call webhooks or use other integrations.
When you create a project, Sentry creates default monitors for it: an Issue Stream Monitor that tracks new issues of all types, and an Error Monitor based on grouping and fingerprint rules. On top of those you can add custom monitors: metric monitors for thresholds on errors, spans, logs, releases and application metrics, cron monitors for scheduled jobs, uptime monitors for HTTP endpoints, and mobile build size monitors.
Older guides and blog posts talk about "issue alerts" and "metric alerts". This guide uses the names in Sentry's current documentation.
Why are my Sentry alerts so noisy?
Three defaults in the Alert builder explain most of the noise:
- Environment defaults to All Environments. Staging, QA and preview deployments page the same people as production.
- Actions run on every issue that matches the filters, and a rule can have no filters. A rule on "a new issue is created" with no If block reports every new issue, including low-priority ones.
- The action interval defaults to "Get notified on every trigger". A busy issue can send a message for each matching event.
The cost of that noise is attention. Google's SRE book sets an upper bound of two incidents per 12-hour on-call shift, because handling one takes about six hours including follow-up (Being On-Call). A small team cannot absorb dozens of pages a day, so it stops reading them, and then misses the real one.
Which Sentry alert triggers should a small team use?
Triggers, the "When" part of a rule, decide which events start an evaluation. They use ANY logic and each one can be added once per rule.
Trigger names as listed in Sentry: Create Alerts, read 13 November 2026. Sentry also lists triggers tied to its Seer AI stages.
One constraint shapes everything else: frequency filters and event attribute filters only work with the "a new issue is created" and "an event or issue activity is captured" triggers. If you want to page on volume, such as users affected in the last 15 minutes, build the rule on one of those two.
How do you filter Sentry alerts so only real problems page someone?
Filters are the "If" part. Each If block uses one qualifier: all, any or none of its filters must match. The filters that do the most work for a small team:
- Environment. Choose production in step 2, or filter the environment tag inside the If block for more complex routing.
- Issue priority. Filter on priority greater than or equal to high, medium or low. Sentry assigns priority from the log level for errors: high for ERROR and FATAL events, medium for WARNING, low for DEBUG and INFO.
- Number of users affected. More than a count in 5 minutes, 15 minutes, an hour, a day, a week or 30 days, or relatively higher than a past interval. Users affected is a better paging signal than raw events, because one broken client can send thousands of events.
- Percentage of sessions affected. Compares the issue's event count with an estimate of project sessions; intervals are limited to 5, 10 and 30 minutes and one hour.
- Issue category and type. Category can be error, feedback, outage, metric, db_query, http_client, frontend or mobile. An outage category filter is how you route uptime problems.
- Issue age and assignment. Useful for a "new and nobody owns it" channel rule.
Pick the threshold from your own data. Open the issues that would have paged last month and set the users-affected number so that only the ones you would have wanted to be woken for pass.
How often should a Sentry alert notify?
The action interval controls how often an alert can run its actions for a given issue. Matching events inside the window do not trigger the actions again. The options are every trigger (the default), 5, 10, 30 and 60 minutes, 3, 12 and 24 hours, one week and 30 days.
For a paging rule, 30 or 60 minutes is enough: the person paged is already working on it, and a second page adds nothing. For channel rules, 3 to 24 hours keeps a long-running issue visible without flooding the channel. Avoid "every trigger" on anything that reaches a person.
Where should Sentry alerts go: Slack, PagerDuty or email?
Send each class of problem to the place where it will be handled, and keep only one class on the pager. The Then block offers:
- Notify on preferred channel: suggested assignees, a team or a member, delivered by each person's own notification settings. Suggested assignees include current assignees, owners from ownership rules and authors of suspect commits. If none is found, the fallback is recently active members by default, or all project members, or no one.
- Chat: Slack (workspace and channel; Sentry suggests adding a channel or user ID to avoid rate limiting), Microsoft Teams and Discord.
- On-call tools: PagerDuty (you pick the service and the severity) and Opsgenie (team and priority).
- Tickets: Jira, Jira Server, GitHub, GitHub Enterprise and Azure DevOps.
- Webhooks and other integrations.
One alert can hold several If/Then blocks, so a single rule can page for high priority and post to a channel for medium. To send each problem to the right people, set up ownership rules that match file paths, URLs, modules or tags. They apply to error issues only. Importing a CODEOWNERS file needs a Business or Enterprise plan, and ownership rules take precedence over code owners.
Sentry metric and uptime alerts: when do you need them?
Issue-based rules catch exceptions. They miss slow pages, failing scheduled jobs and a site that does not answer at all. That is what monitors are for:
- Metric monitors track a dataset and metric at an interval, with an absolute threshold, a percentage change, or dynamic anomaly detection. Sentry suggests fixed or percentage thresholds when you know what good and bad look like, with examples such as a crash rate above 1% or transaction volume dropping to zero.
- Uptime monitors check a URL from several regions. To avoid false positives, an issue is only created after three consecutive failures by default. Sentry automatically creates one for the most frequently seen hostname in your error data.
- Cron monitors track scheduled jobs with a schedule, a check-in margin, a maximum runtime and a failure tolerance.
The trap: an uptime monitor creates an issue but does not notify on its own. Sentry's docs say to configure an alert that matches outage issues and add the actions you want. Without that rule, your downtime detection writes to a list nobody is watching.
Sentry alert rules for a small team: a starter set
Five rules cover most small products. Create them, delete or disable the catch-all rules you had, and run the set for two weeks before you tune it.
Our recommended starting point. Every value in the table is a setting the Sentry docs describe; the thresholds are yours to pick from your own traffic.
Rule 1 is the only one that pages. If it fires more than a couple of times a week, raise the threshold or fix the underlying issue; a page that is regularly ignored is worse than no page. If you deploy several times a day, a release-based rollback path makes the page shorter to handle; our guide to blue-green deployment covers one way to get it.
How do you test Sentry alerts before you rely on them?
Each If/Then block has a Send Test Notification button that sends a test event through that block's actions, which confirms the integration is wired up. Sentry notes that event-based triggers may behave differently in a real run, so test the logic too:
- Throw a test exception from a staging build and confirm that no production rule fires.
- Throw it from production with a test tag, confirm it reaches the channel, then resolve it and throw it again to check the regression rule.
- Point an uptime monitor at a URL that returns an error, wait for three failed checks, and confirm the outage rule pages.
- After two weeks, list every alert that fired and mark which ones someone acted on. Remove or demote the rest.
For a wider pre-launch review, our production readiness checklist includes monitoring and alerting checks alongside security and deployment.
What does a good Sentry alert setup look like?
What good looks like
- ✓Only one rule reaches the pager, and it is limited to production.
- ✓Every rule has an environment filter and a deliberately chosen action interval.
- ✓Paging uses users affected or percentage of sessions rather than raw event counts.
- ✓New and regressed issues go to one team channel with suggested assignees.
- ✓Low-priority issues become tickets instead of messages.
- ✓Uptime and cron monitors have an alert that matches their issues.
- ✓Ownership rules route errors to the team that owns the code.
- ✓Every rule has been tested, and the set is reviewed monthly against what people actually acted on.
Alerts are one part of operating a service. For the rest, see our DevOps best practices guide.
Want someone to own the alerts at 3 a.m.?
Valletta Software Development runs DevOps and SRE support on a follow-the-sun rotation across five timezone clusters. A named engineer responds to a P1 within 15 minutes and to a P2 within 30, and targets are written into the contract for each engagement. We start by tuning the alerts you already have. DevOps support is priced from $45 per hour.
FAQ
How do I set up Sentry alerts?
In Sentry, go to Monitors, then Alerts, and create an alert in five steps: select a source, filter by environment, build the When, If and Then blocks, set throttling, then name and save. Start with one paging rule for production and one channel rule for new and regressed issues, as the docs describe.
What are Sentry alerts best practices?
Sentry recommends alerts that are valuable and not too noisy: use filters to narrow issues to those that matter, select only the triggers you need, and match routing to how your team triages. In practice that means production only, priority or frequency filters, a set action interval, and separate pages from channel notifications.
Why am I getting too many Sentry alerts?
Usually because a rule covers all environments, has no filters, and notifies on every trigger, which is the default action interval. Each matching event then sends a message. Restrict the rule to production, add a priority or frequency filter, set an interval such as 30 minutes or 3 hours, and move low-priority issues to tickets.
Can Sentry send alerts to PagerDuty or Slack?
Yes. With the integrations installed, an alert's Then block can notify Slack, Microsoft Teams, Discord, PagerDuty, Opsgenie or a webhook, and can create Jira, GitHub or Azure DevOps tickets. For PagerDuty you pick the service and the severity; for Slack, the workspace and channel. Details are in Create Alerts.
What is the difference between Sentry monitors and alerts?
Monitors decide when errors, metrics, cron jobs or HTTP checks become issues; Sentry creates default monitors for each new project. Alerts act on those issues when they match your triggers and filters, by notifying people, creating tickets or calling integrations. A monitor without an alert creates issues but does not notify anyone.
How do I stop duplicate Sentry notifications for the same issue?
Set the action interval on the rule. It controls how often the alert can run its actions for one issue: options range from every trigger through 5, 10, 30 and 60 minutes, then 3, 12 and 24 hours, one week and 30 days. Repeats inside the window do not notify again.