mc keeps one --resolve IP per host, and Coolify always pinned the IPv4
address of a dual-stack endpoint. On servers without IPv4, uploads failed
with "network is unreachable". A helper container on the upload network
now connects to each validated IP, IPv4 first, and the first IP that
connects is pinned. S3 restores are now pinned in the same way.
Closes#12397
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Unset the helper's build secret variables before Compose pulls or starts
services, so containers get the runtime .env values. Give build secrets
the raw stored value instead of the quoted .env form.
Closes#12409
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Docker Compose applications dropped the stored SERVICE_URL_<NAME>_<PORT> and SERVICE_FQDN_<NAME>_<PORT> variables and only generated the forms without a port, so compose references to them were blank. Generate the port-suffixed forms from the current domain for runtime and build-time, production and preview deployments.
Closes#12427
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Every channel has a block of its own, every version after this one reads only these blocks
- Every stable sync also fills the coolify and traefik keys older instances read
- An instance keeps a copy of versions.json and reads upstream.json of the image until it has one
- Adds the channels stable, preview, nightly and custom
- Preview and nightly read /coolify/release.json on the CDN and take the newest version of their channel and the more stable ones, nightly gets a new build of next or the release branch twice a day
- Custom installs any image tag of ghcr.io by hand, such as pr-1234 for the build of pull request 1234
- After a switch to a more stable channel, the instance keeps the updates of the channel it left until the stable release of its version is out
- The local versions.json moves to storage/app and is replaced in one step, a channel change replaces it without the version of the old channel
Save build_server_id as soon as a build server is selected, so the cancel
cleanup finds the helper container on the build server. Let the final
helper stop run after a cancel.
Closes#12383
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Two applications that build the same Docker image name push the same
tags (commit SHA, latest, pr-<id>), so one can run the image of the
other. Deployments, the General form, and the API now reject such a
name for applications of the same team.
Closes#9834
ServerManagerJob and the Docker cleanup step of ScheduledJobManager
load the settings of every server with SELECT *. That reads the
patch_check_result JSON, which holds the full update list of each
server. With 12,709 rows, the scan took about 4 ms, but the full fetch
took 390-536 ms and 294 MB of PHP memory. Without the column it took
74 ms and 111 MB.
Add ServerSetting::withoutPatchCheckResult() and use it in both loops.
Queued jobs reload the full settings through SerializesModels.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
An OR of two correlated EXISTS subqueries made Postgres estimate a
per-server scan of the runner executions, so JIT compilation took most
of the time of the per-minute reconcile query. Select the server ids
from both runner tables with one UNION and index executions by status
and server.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Public pull request previews no longer get production values, and no
preview falls back to production variables. Existing previews were
deployed with these values; for example, their databases were
initialized with the production password, so they broke on redeploy.
Previews that exist at the upgrade get keeps_production_values and
keep the earlier behavior. Previews created later do not. Server
transfer moves the flag.
- Add an Axiom edge domain setting (default us-east-1) and use the /v1/ingest endpoint, so EU datasets work
- Remove the Highlight drain (service ended) and stop its running containers
- Update Fluent Bit from 2.0 to 5.1 and use the warn log level
- Restart running drains when the generated files change, with the new image pulled before the old container stops
- Show log drain validation errors under their fields
Closes#9898
Proxies started before v4.4.4 kept the old nginx config after an upgrade:
the upstream IP was resolved only once and IPv6 connections were reset.
- Label new proxies with a configuration version
- Status checks (Sentinel and SSH) recreate a running proxy with an
older or missing version once, at most every 10 minutes
- Remove the old proxy only after the image pull, so a failed pull
keeps it running
Closes#10198
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Add a "Use path-style URLs" setting to S3 storages. It is detected from
the endpoint (Alibaba OSS, Tencent COS, Volcengine/BytePlus TOS), and the
connection test switches to virtual-hosted style when the provider
rejects path-style requests. Backups pass `--path off` to mc.
Closes#10347
- The restart limit claim sent only one stop. If that stop failed or came
too late, the container kept restarting past the max restart count.
- Sentinel pushes and status checks now dispatch the stop again while a
container still runs; the notification is still sent once.
Closes#10761
The compose parser queued ServerFilesFromServerJob for every volume on
each parse, and pages that only show data parse the Compose file. Each
sync also sent two SSH commands for each file.
- Queue the sync only for a compose deployment, and for each service
resource with file storages when the service starts or deploys.
- Read the compose service names without a parse on the Domains and
scheduled task pages.
- Check all file storages with one remote command, and stop when a
symbolic link, a special file, or an unknown state is at a file path.
- Keep the application path in a file storage when a pull request
preview is parsed. Preview deployments of the current parser write
their `-pr-<id>` copies through LocalFileVolume::forPullRequest().
- Remove the unreachable file sync from the service resource page.
Sending SSH command batches on stdin changes how every remote command
reaches a server, which is too risky for a patch release. Move it to
next and restore the heredoc transport from before 1a2a5362a. The
shared test helpers stay because they work with both transports.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Bitbucket Cloud has no pull request refs over git, so a redeploy without
a webhook commit checked out the base branch. Reuse the latest known
commit of the pull request, or fail the deployment when none is known.
Closes#10806
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
peaklabs-dev
Andras BacsaiandClaude Opus 5.5
github-actions[bot]
