PricingBlogDocumentationDocsSupport
LoginRegister

GlitchTip 7.0 released

GlitchTip 7.0 is out! It brings simpler permissions, much better stack traces, and a batch of quality of life improvements. Make sure to read Breaking Changes to understand how these changes affect you!

Simpler Permissions

In 7.0, we've reworked the old permissions scheme into one that is focused on roles. For both users and their API tokens, what you can do depends on your role in a given Organization. As a reminder, GlitchTip's roles from least to most access go as follows:

Member < Admin < Manager < Owner

Please consult this table of changed behaviors to see how this might effect you. We suggest re-organizing users in your organization to fit this scheme.

Action Before 7.0 7.0 and later
Delete an organization Manager or above Owner
Rename a team Any member Admin+
Update a project Any member Admin+
Create a client key (DSN) Any member Admin+
Hide an environment Any member Admin+
Create a project alert Admin or above, or a member on one of the project's teams Admin+
Update or test a project alert Any member Admin+
Create, edit or delete an uptime monitor Any member Admin+
Create a status page Any member Admin+
Grant, invite, change or remove an owner Manager or above Owner

You cannot update any of the API tokens you've already created, so some new tokens will have to be minted to adapt to the updated behavior. Furthermore, some endpoints have had gates added to them, and API tokens will have to be created to access them:

  • org:read to view uptime monitors and status pages
  • org:read or org:admin for billing
  • event:read for issues

Better Stack Traces

Symbolication turns a minified, compiled or obfuscated stack trace back into real function names, files and line numbers, using the sourcemaps, debug files and ProGuard mappings you upload. We rebuilt it in 7.0.

  • Less memory. GlitchTip used to re-read your uploads for every event, which caused memory spikes on busy installs. Now each upload is processed once into a cache that later events reuse. Nothing to configure.
  • Android stack traces are deobfuscated. ProGuard mappings now work, including ones uploaded with sentry-cli upload-proguard 3.x, which used to be rejected while the CLI reported success. Frames also show their real line numbers.
  • Better native stack traces. Windows crashes now match their uploaded PDBs and get source context from source bundles. Every architecture of a universal (fat) binary is used, so arm64 crashes symbolicate. Caller frames point at the line of the call, not the line after it.

Alerts

  • Alerts by environment. A project alert can now name the environments it applies to, so a flood of errors from staging no longer pings you about production. Leave it empty and the alert matches every environment, as before.
  • Slack alerts. A new Slack alert recipient posts formatted messages to a Slack incoming webhook, for both issue and uptime alerts.

AI Assistants

GlitchTip's built-in MCP server lets AI assistants read your issues, events, performance data and logs. In 7.0 it returns a trimmed event by default to save on token usage. It always keeps the exception and the frames where the crash happened, and a new get_event_detail tool fetches anything left out.

Other Improvements

  • Issue lookup by short ID. Search for a short ID like BACKEND-1A in the issue list and go straight to that issue.
  • Sortable uptime monitors. Sort your monitors by name, creation date, project or status.
  • Clearer dialogs and buttons. Deleting something now always asks you to confirm, and buttons make it clearer what each choice does.
  • Faster first load. A cold first load of the app drops from 126 KB to 31 KB.
  • Tunneled events accepted. If you send events through a proxy, like the JavaScript SDK's tunnel option, GlitchTip now reads the DSN from the envelope instead of rejecting it with a 403.
  • Lighter ingest. All event ingest now runs on our Rust pipeline, which uses less memory and CPU. The GLITCHTIP_RUST_INGEST variable is now ignored.
  • AWS RDS IAM. Self-hosters can authenticate to AWS RDS with IAM tokens instead of a database password. Turn it on with DATABASE_IAM_AUTH=true and sslmode=verify-full.
  • Django 6.1. Security support through December 2027.

Breaking Changes

Hosted users only need to read For All Users. Self-hosting? Read both sections.

For All Users

  • All-digit org slugs renamed. To improve compatibility with third-party API clients, every URL with an organization in it now also accepts the organization's numeric id, so slugs can no longer be all digits. Existing all-digit slugs got the first free "-N" appended: 2024 becomes 2024-1, or 2024-2 if that one is taken. If your slug was all digits, update your links, bookmarks, CI configs and API scripts.
  • Some tokens need reminting. An API token with scopes beyond its owner's role stops working. Mint a new one.
  • Some actions now need admin. Members lose a few configuration actions. See the table in Simpler Permissions.

For Self-Hosted Users

  • Upgrade from 6.x. Any 6.x release can upgrade to 7.0. On 5.x? Upgrade to 6.x first.
  • PostgreSQL 15 or newer is required. The easiest path is pg_dump and restore onto a new server. Upgrade Postgres while you are still on GlitchTip 6.x.
  • Renamed environment variables. 7.0 no longer reads the old names in the table below. It logs a warning and uses the default. If you still use an old retention name, retention falls back to 90 days and older events will be deleted.
  • Check TRUSTED_PROXIES. The default changed from * to private, so X-Forwarded-For is only read from proxies on private addresses. Behind Cloudflare, set TRUSTED_PROXIES=cloudflare,private. Behind another proxy with public addresses, list its ranges plus private. Until you do, users behind that proxy share one login rate limit, and GlitchTip logs a warning. TRUSTED_PROXIES=* restores the old behavior.
  • uWSGI removed. The legacy uWSGI server is no longer installed in the Docker image. If you override GlitchTip's startup command, we recommend switching back to the default, bin/start.sh.
  • No more memory-based restarts. GlitchTip no longer restarts workers when memory grows. Reported memory will look higher because those restarts were hiding the spikes. Size your containers for spikes. For periodic restarts, set GRANIAN_WORKERS_LIFETIME.
  • Removed settings. DATABASE_CONN_MAX_AGE, USE_ASYNC_SERVER, IS_WORKER, GLITCHTIP_EMBED_WORKER and SERVER_ROLE no longer do anything and log a warning if set.
Retired name (now ignored) Use instead
GLITCHTIP_MAX_EVENT_LIFE_DAYS* GLITCHTIP_RETENTION_DAYS
GLITCHTIP_MAX_TRANSACTION_EVENT_LIFE_DAYS* GLITCHTIP_TRANSACTION_RETENTION_DAYS
GLITCHTIP_MAX_UPTIME_CHECK_LIFE_DAYS* GLITCHTIP_UPTIME_RETENTION_DAYS
GLITCHTIP_MAX_FILE_LIFE_DAYS* GLITCHTIP_FILE_RETENTION_DAYS
GLITCHTIP_LOGS_COLD_DAYS* GLITCHTIP_LOG_RETENTION_DAYS
GLITCHTIP_EVENTS_HOT_DAYS GLITCHTIP_EVENT_HOT_DAYS
GLITCHTIP_LOGS_HOT_DAYS GLITCHTIP_LOG_HOT_DAYS
ENABLE_OPEN_USER_REGISTRATION** ENABLE_ORGANIZATION_CREATION

* marks a retention setting. If you have the old name set, update it.

** controls organization creation and needs one extra step. If you set ENABLE_OPEN_USER_REGISTRATION, set ENABLE_ORGANIZATION_CREATION to the same value before upgrading and remove the old name. In 6.x the old name overrode the new one, so if you set both to different values, 7.0 will flip your organization-creation setting.

What's Next

While this is the culmination of a lot of work, there's still plenty more on the horizon. Long term, we're working on finalizing an on-demand billing system. On the smaller side, you can look forward to Organization SSO Login and tweaks to uptime, issues and alert behavior, just to name a handful.

Support Us

GlitchTip is 100% built and funded by users. No VC. We have no one to answer to but you.

Support the project

Community & Social