How to migrate from Sentry to GlitchTip

Import the Sentry organisation, move each DSN one service at a time, point source map uploads at GlitchTip, rebuild alerts and cron checks, then close Sentry.

Start with one workloadThe steps

Written for
The engineer who will do the work
Or run by Pilae
At a fixed price, then operated

In short

  • GlitchTip accepts events from the Sentry SDKs, so each application keeps its code and changes the DSN it reports to, one service at a time.
  • GlitchTip’s import command copies the Sentry organisation, members, teams and projects with their client keys; issue history and alert rules stay behind.
  • Session Replay, profiling, release health and Crons check-ins are ignored at ingest, and heartbeat monitors take the place of cron checks.
  • Source map uploads move to glitchtip-cli or keep sentry-cli with its URL pointed at GlitchTip, and a deliberate error in staging proves the stack trace resolves.
On this page (9)

Moving from Sentry to GlitchTip is a re-pointing of clients, not a data migration. GlitchTip accepts events from the Sentry SDKs, so no application changes its instrumentation: each one changes the DSN it reports to. What moves around that is configuration. An import command copies the organisation, its people and projects; source map uploads in CI are pointed at the new host; alert rules and cron checks are rebuilt by hand. Issue history stays in Sentry. This guide covers the order we run it in, and the places it usually goes wrong.

Why teams move from Sentry to GlitchTip

Sentry is a strong product, and the teams that leave it usually do so for three reasons.

  • What an error event carries. A stack trace comes with URLs, headers, user identifiers, breadcrumbs and sometimes the values of local variables. With GlitchTip on your own machines, those events are stored in the country you choose, and ingest scrubbing can redact secrets and card numbers before anything is written.
  • A smaller system to run. GlitchTip is a Django application on PostgreSQL, with Valkey as an optional cache and queue. Self-hosted Sentry, the closer match on features, runs Kafka, ClickHouse, Snuba and Relay among dozens of containers.
  • A licence without conditions. GlitchTip is MIT-licensed, with no paid edition and no bill per event once you run it yourself. Sentry’s server is source-available under the Functional Source License, which allows internal use.

What you give up is real. Session Replay, profiling and release health have no counterpart and are ignored at ingest. Crons check-ins give way to heartbeat monitors. Alerts are simpler: GlitchTip counts events per issue, with no metric alerts on error rates or response times. If your teams rely on replay or profiling, self-hosted Sentry is the closer fit. And someone has to run GlitchTip, which is the part Pilae takes on.

Step 1: list projects, DSNs and alerts in Sentry

The import brings the structure across. Everything around it has to be listed first, and Sentry’s API does most of the listing. Create a personal token from the Personal Tokens page under the account menu, with read access to the organisation, its projects, members, teams and alerts.

  1. Projects: GET /api/0/organizations/{org}/projects/ lists every project.
  2. Client keys: GET /api/0/projects/{org}/{project}/keys/ returns each project’s keys with their DSNs. Record which DSN each service and environment uses today; it is the list step 4 works through.
  3. Alerts and monitors: on sentry.io, the Monitors & Alerts API lists alerts with Fetch Alerts (GET /api/0/organizations/{org}/workflows/) and monitors with Fetch an Organization’s Monitors (GET /api/0/organizations/{org}/detectors/). Each alert comes back with its triggers, conditions and actions. Cron monitors are in the Crons API, under Retrieve Monitors for an Organization.
  4. Teams and members: List an Organization’s Teams and List a Team’s Members give the ownership the import should reproduce.
  5. In the code and CI: where each service sets its DSN, its sample rates, which SDK features it turns on (tracing, Session Replay, profiling, cron check-ins), and every pipeline step that runs sentry-cli or a Sentry bundler plugin with a SENTRY_AUTH_TOKEN.
  6. Integrations: the Slack workspaces, issue trackers and webhooks Sentry posts to.
Sentry GlitchTip
Organisation, teams, members Organisation, teams, members, through the import command
Project and client keys (DSN) Project and client keys, IDs kept by the import command
Errors and issues Errors and issues
Tracing Performance: transactions and their spans
Logs Logs, from the Sentry SDK or over OpenTelemetry
User Feedback User reports
Releases and deploys Releases and deploys, through glitchtip-cli or sentry-cli
Source maps and debug files Source maps, artifact bundles with debug IDs, debug files
Issue alerts Project alerts: a number of events per issue in a timespan
Metric alerts No equivalent
Uptime monitors Uptime monitors: Ping, GET, POST, TCP port and SSL
Crons Heartbeat monitors
Release health (sessions) Ignored at ingest
Session Replay and profiling Ignored at ingest
Issue history Stays in Sentry

Step 2: import the organisation into a scratch instance

GlitchTip’s import management command reads Sentry’s API and creates the organisation, its members with user accounts, its projects with their client keys, and its teams with their members and projects. Run it as a server administrator, in the GlitchTip container:

# On a scratch instance first. The token is a Sentry personal token
# with read access to the organisation; keep it out of shell history.
./manage.py import https://sentry.example.com "$SENTRY_TOKEN" acme

The arguments are the Sentry address, the token and the organisation slug. The command first checks the token against the root of Sentry’s API and stops with “Bad auth token” if it is not tied to a user. Run this way, each project keeps its ID and its client keys, so a DSN should change only in its host. GlitchTip’s changelog points to this command for exactly that reason: organisation admins can also import through GlitchTip’s API, but projects imported that way get new IDs and keys, so every DSN changes.

Two limits shape how we use it. Upstream verified the command against an older self-hosted Sentry, not sentry.io, so it runs first on a scratch instance, and we compare the members, projects, keys and team memberships it created with the lists from step 1 before running it for real on an empty production instance. Earlier GlitchTip releases could not bring members across with this command, so run it on a current release and check that every member is there. And it copies structure only: no events, issues, alert rules, releases, source maps, integrations or passwords. If your Sentry is self-hosted on a private network, GlitchTip refuses to call internal addresses unless GLITCHTIP_ALLOW_PRIVATE_IPS is set.

Step 3: set up sign-in, alerts and monitors

Sign-in. GlitchTip signs people in through django-allauth, and its documentation names Keycloak as an identity provider to configure through OpenID Connect. In the Django admin, under /admin/socialaccount/socialapp/, add a social application with the provider OpenID Connect, a provider ID, the name people will see on the button, the client ID and secret from your identity provider, and in Settings a JSON object whose server_url is the provider’s discovery URL. The callback to register in Keycloak is allauth’s, /accounts/oidc/<provider-id>/login/callback/ on your GlitchTip host. Then close self-registration with ENABLE_USER_REGISTRATION=False, and set ENABLE_SOCIAL_APPS_USER_REGISTRATION=True if colleagues who were not in Sentry should get an account on their first sign-in through the identity provider: by default it follows the first setting. Imported accounts carry an email address and a name, never a Sentry password, so check with one imported account that its first sign-in through the identity provider lands on that account, with its teams, before telling everyone.

Alerts. A GlitchTip project alert fires when an issue receives at least a set number of events within a set number of minutes, once per issue, optionally only for the environments you name. It can also fire when one of the project’s uptime monitors fails. Rebuild each Sentry alert from the step 1 list in those terms. Recipients can be email, Slack, Discord, Microsoft Teams, Google Chat, Zulip, ntfy, Feishu or a generic webhook, which covers Mattermost and Rocket.Chat. An alert with no recipient sends email, so the instance needs a working mail relay from the first day. Webhooks to internal services, such as a Mattermost on your network, need GLITCHTIP_ALLOW_PRIVATE_IPS, because GlitchTip blocks private addresses by default.

Monitors. Sentry’s uptime checks become GlitchTip uptime monitors, of type Ping (a HEAD request), GET or POST against a URL, or TCP Port or SSL. Sentry Crons become heartbeat monitors: GlitchTip shows a URL when the monitor is created, and the job sends a POST to it at the end of each successful run. The SDK’s own check-ins are ignored, so remove them from the job rather than leave them looking like they work. Link each monitor to a project whose alerts have uptime notifications switched on, or nobody hears that it went down.

Step 4: move the DSNs one service at a time

A Sentry SDK is initialised with one DSN, so an event goes to one place. The parallel run therefore moves by environment and by service rather than sending everything twice: staging first, then production one service at a time, each one moving with its alerts already in place in GlitchTip.

# Before: the DSN Sentry issued for the project
SENTRY_DSN=https://<public-key>@o<org-id>.ingest.sentry.io/<project-id>
# After: GlitchTip's host; check the key and project ID in the project's settings
SENTRY_DSN=https://<public-key>@errors.acme.ch/<project-id>
  • Change configuration, not code. Server-side SDKs, Python’s among them, read the SENTRY_DSN environment variable when no DSN is set in code. Where it is hard-coded, move it into configuration on the way.
  • Set sample rates first. GlitchTip’s SDK documentation advises a low trace sample rate in production to save disk space. Lower traces_sample_rate or tracesSampleRate before production traffic arrives, not after the first bad deploy, and switch off what GlitchTip ignores: its own JavaScript setup turns off session tracking.
  • Browsers and mobile apps report from the internet, so GlitchTip’s ingest endpoint has to be reachable from there. A mobile build carries its DSN inside it, so versions already installed keep reporting to Sentry until people update.
  • Check each move with a real event in GlitchTip, from the service itself or with glitchtip-cli send-event.

Step 5: point source map uploads and releases at GlitchTip

Without source maps, a minified JavaScript stack trace is unreadable, so every pipeline that uploads to Sentry changes in the same release as the DSN. GlitchTip reads Sentry’s artifact bundles and debug IDs, so with the debug IDs injected before the upload, source maps resolve without a release name.

  • Keep sentry-cli. Set SENTRY_URL to the GlitchTip host, as Sentry documents for self-hosted servers, and SENTRY_AUTH_TOKEN to a GlitchTip token from Profile, Auth Tokens, with the project:releases scope. SENTRY_ORG must be the organisation slug exactly as GlitchTip shows it: the upload routes match it case for case.
  • Or switch to glitchtip-cli, GlitchTip’s own MIT-licensed tool, which the project marks as beta. It reads the same SENTRY_* variables and .sentryclirc files, and covers source maps, releases, deploys and debug files.
  • Bundler plugins, such as Sentry’s Vite or webpack plugin, take the GlitchTip host in their url option or from SENTRY_URL.
# CI step after the build
export SENTRY_URL=https://errors.acme.ch/
export SENTRY_AUTH_TOKEN="$GLITCHTIP_TOKEN" SENTRY_ORG=acme SENTRY_PROJECT=web-frontend
glitchtip-cli releases new "$GIT_SHA"
glitchtip-cli sourcemaps inject ./dist
glitchtip-cli sourcemaps upload ./dist --release "$GIT_SHA"
glitchtip-cli releases finalize "$GIT_SHA"

Then throw a deliberate error from staging and check that GlitchTip shows the original file and line. Once every pipeline has moved, revoke the Sentry tokens stored in CI.

Step 6: run both, then close Sentry

From the first DSN to a closed organisation

  1. Staging: Staging environments report to GlitchTip first, and source maps are proven there.
  2. Services: Production services move one by one, each triaged in GlitchTip from the day its DSN changes.
  3. Stragglers: Sentry projects that still receive events show what has not moved, such as an old mobile build or a forgotten job.
  4. Lookups: Sentry stays open for reading old issues until the last useful one has aged out.
  5. Close: Integrations and tokens are removed, the subscription cancelled and the organisation deleted.

Before the organisation goes, export anything you want to keep: Sentry’s API lists a project’s issues, and that list is the only record of them once Sentry deletes the data. Deleting a Sentry organisation does not cancel its subscription, so cancel first, under Settings, Subscription, Manage Plan, Cancel Subscription. Then an owner removes the organisation with Remove Organization at the bottom of its settings page. Sentry emails the scheduled date, the organisation can be restored until then, and after it the deletion is irreversible.

Running GlitchTip in production

GlitchTip is light, but a production instance still needs care in a few places.

  • PostgreSQL holds issues, events, users and uptime checks, partitioned by date, so expiring old events drops a partition instead of running a long delete.
  • Valkey is optional. Without it, PostgreSQL carries the task queue, cache and sessions, more slowly.
  • Object storage for source maps, debug files and, with cold storage on, events archived as Parquet files.
  • Retention, agreed before the first event: it decides the disk.
  • A mail relay, because alerts and invitations go out by email.
  • Upgrades read from the changelog. Migrations run on start, and releases rename settings: a retired retention variable is ignored with a warning, retention falls back to its default, and the next maintenance run deletes anything older.

Pilae operates GlitchTip this way: PostgreSQL with Valkey beside it and a bucket in the same region, ingest scrubbing on, sign-up closed and sign-in through Keycloak or your own identity provider over OpenID Connect. Only the ingest endpoints are published through the gate, for browsers and mobile apps, and the interface answers on your private network. PostgreSQL and the bucket are backed up together daily, encrypted, to an offsite location in the country your contract names, and on Enterprise a restore test every month is recorded in the console. Probes check the ingest endpoint every 60 seconds. The Pilae Agent tries each release on a copy and applies it in your window once you approve. It runs on your premises or on dedicated machines in Pilae Cloud.

Let Pilae run the move

Pilae runs Sentry to GlitchTip migrations at a fixed price: the inventory, the organisation import and its check, sign-in through your identity provider, alerts and monitors rebuilt, the DSN move service by service, the CI changes for source maps, the parallel run and the closing of the Sentry organisation. Then we operate GlitchTip for you under a contract that, from the Business plan, includes a monthly availability commitment with service credits. See migration services for what the fixed price covers, or tell us which services report to Sentry today and we will scope it.

Questions

Do our applications need new SDKs?

No. GlitchTip accepts events from the Sentry SDKs, so each application keeps its code and changes the DSN it reports to. Errors, performance transactions, logs and user feedback are stored. Session Replay, profiling, release health sessions and Crons check-ins are accepted and ignored, so features that depend on them stop showing data.

Does our issue history come across?

No. GlitchTip's import command copies the organisation, its members, teams and projects with their client keys, and nothing else: no events, no issues, no alert rules, no releases or source maps. Old issues stay readable in Sentry for as long as you keep the organisation and its retention allows.

Do our DSNs change?

The host does. The import command keeps each project's ID and copies its client keys, so a DSN should change only in its host name. Upstream verified the command against an older self-hosted Sentry, not sentry.io, so we run it on a scratch instance first and compare every project's keys with Sentry's before any DSN moves.

What replaces Sentry Crons?

Heartbeat monitors. GlitchTip ignores the check-ins the Sentry SDKs send, so each scheduled job sends a POST to the heartbeat URL GlitchTip shows for its monitor instead, and GlitchTip marks the monitor down when the request stops arriving.

Is GlitchTip open source?

Yes. Backend and frontend are MIT-licensed, with no paid edition and no feature held back from a self-hosted install, and glitchtip-cli is MIT too. Burke Software and Consulting, which maintains GlitchTip, asks for-profit companies running it in production to buy a support licence. That buys priority support and invoices, not features, and the licence does not require it.

How long does a Sentry to GlitchTip migration take?

It depends on the number of services and where each one sets its DSN, the CI pipelines that upload source maps, and the alert rules to rebuild, not on event volume. The inventory in step 1 settles it. Pilae quotes the move at a fixed price once it is done.

Let us run the move for you

An inventory, a parallel run and a cutover plan, at a fixed price. Then we operate the app.

Start with one workload

  • Go deeperSaaS migrationA fixed-scope migration off Google Workspace, Microsoft 365, Zapier, Airtable and ChatGPT to open-source apps, with a parallel run and a cutover plan.
  • CheckManaged operationsPatching, daily encrypted backups, monitoring and incident response for your open-source apps, under an SLA with an availability commitment of up to 99.9%.