Compare commits

...

3 Commits

Author SHA1 Message Date
Nagesh Bansal
67418bb132 docs: fix the signoz docs troubleshooting guide url 2026-06-25 07:42:57 +05:30
Nagesh Bansal
2e2517449d docs(deploy): title migration guide for both install script and deploy 2026-06-25 06:52:36 +05:30
Nagesh Bansal
0150c55361 docs(deploy): restructure migration guide and add Docker Swarm 2026-06-25 06:38:48 +05:30

View File

@@ -1,48 +1,76 @@
# Migrating from the install script to Foundry
# Migrating from the install script and `deploy/` to Foundry
The install script (`install.sh`) and the bundled Compose and Swarm files
under `deploy/` are deprecated in favor of [Foundry][foundry], the supported
way to install and manage SigNoz. This guide moves an existing Docker Compose
or Docker Swarm deployment to Foundry and reattaches your existing volumes, so
your data is preserved.
> [!IMPORTANT]
> The install script is now deprecated and will no longer receive updates.
> This guide is only for **existing** `install.sh` / `deploy/` deployments.
> Setting up SigNoz for the first time? Skip migration and install Foundry
> directly: [SigNoz install docs][install-docs].
This guide walks you through migrating an existing SigNoz deployment running via
Docker Compose to [Foundry](https://signoz.io/docs/install/docker/).
## How it works
> [!NOTE]
> Setting up SigNoz for the first time? You don't need this guide — follow the [SigNoz installation docs](https://signoz.io/docs/install/) instead.
Foundry splits a deployment into two commands:
## Overview
To stay up to date on new installation platforms and patterns, please refer to [Foundry](https://github.com/SigNoz/foundry).
- `foundryctl forge` generates the deployment manifests from a `casting.yaml`.
It never touches running containers, so it is safe to re-run while you
iterate.
- `foundryctl cast` applies those manifests: it (re)creates the containers and
reuses the volumes you point it at.
Two `foundryctl` commands are used throughout this guide:
- **`forge`** — generates deployment manifests from your `casting.yaml`. It does not touch running containers, so it is safe to re-run while you iterate.
- **`cast`** — applies the generated manifests: it creates and starts the containers (and pulls new images).
You write one `casting.yaml`, point a few patches at your existing data
volumes, then cast. The steps below are the same for Compose and Swarm; they
differ only in the casting (step 3) and how you stop the old stack (step 5).
## Prerequisites
- [ ] Install Foundry - `curl -fsSL https://signoz.io/foundry.sh | bash`
## Migration Steps
> [!WARNING]
> **Before proceeding, back up both:**
> - **Your docker volumes** — these hold your data.
> - **Your existing `docker-compose.yaml` (and any config it references)** — keep a copy somewhere safe. The compose manifests are no longer distributed by SigNoz, so this backup is your only way to roll back to your previous setup.
- An existing SigNoz deployment from `install.sh` or `deploy/` (Compose or
Swarm).
- `foundryctl` (installed in step 1).
1. Make a note of the volume names used by your existing deployment for the following components:
- ClickHouse
- SigNoz
- ZooKeeper
## Migrate
> If you used the docker compose file we provided, the volumes will be `signoz-clickhouse`, `signoz-sqlite`, and `signoz-zookeeper-1`.
### 1. Install Foundry
2. Generate your `casting.yaml`. Based on internal testing, the following casting should generate the manifests that mimic the legacy docker compose setup (compare against your backed-up `docker-compose.yaml`). Once created, run `foundryctl forge -f casting.yaml`.
```bash
curl -fsSL https://signoz.io/foundry.sh | bash
```
Foundry has a [Docker Compose example](https://github.com/SigNoz/foundry/tree/main/docs/examples/docker/compose). Please use it as a reference.
### 2. Keep your rollback path
> [!WARNING]
> If your deployment had more than 1 shard or replica, you will need to adjust your manifest volumes accordingly.
This migration reattaches your existing volumes in place; it does not move or
delete your data. The only destructive action is passing `--volumes` / `-v`
when you stop the old stack (step 5), so avoid that flag.
> [!IMPORTANT]
> The `replica` and `shard` macros below are placeholders. Replace them with the values from your existing ClickHouse configuration (check the `macros` section of your current ClickHouse config, e.g. `config.xml`/`metrika.xml`), otherwise the generated manifests will not match your existing data.
> Keep a copy of your existing `docker-compose.yaml` / stack file (and any
> config it references). SigNoz no longer distributes these files, so this copy
> is your only way to roll back.
### 3. Write your `casting.yaml`
Use the casting for your deployment. Both reproduce the legacy single-node
setup (ClickHouse + ZooKeeper + SQLite) and reattach your existing volumes;
they differ only in `spec.deployment.flavor` and the volume-reuse patch
(Compose volumes have a `name` to replace; Swarm volumes are bare, so the whole
entry is replaced). If your deployment ran more than one shard or replica,
adjust the volume patches accordingly. The
[Docker Compose example][compose-example] is a useful reference.
> [!IMPORTANT]
> The `replica` and `shard` macros are placeholders. Replace them with the
> values from your existing ClickHouse config (the `macros` section of
> `config.xml` / `metrika.xml`), or the generated manifests will not match your
> existing data.
<details>
<summary><b>Docker Compose</b> casting.yaml</summary>
```yaml
# casting.yaml
apiVersion: v1alpha1
kind: Installation
metadata:
@@ -61,8 +89,8 @@ spec:
data:
config-0-0.yaml: |
macros:
replica: "example01-01-1" # replace with your existing ClickHouse replica macro (see legacy configuration files for reference)
shard: "01" # replace with your existing ClickHouse shard macro (see legacy configuration files for reference)
replica: "example01-01-1" # replace with your replica macro
shard: "01" # replace with your shard macro
patches:
- target: "deployment/compose.yaml"
operations:
@@ -80,50 +108,165 @@ spec:
value: root
```
> [!NOTE]
> The `user: root` patch on the ZooKeeper service is required so the container can read/write the data in your reused ZooKeeper volume, which was created with `root`-owned files by the legacy compose setup. Without it, ZooKeeper may fail to start with permission errors.
</details>
If you had custom configurations for features like SMTP or additional ingestion processors/receivers, you will need to include those in your casting file via [patches](https://github.com/SigNoz/foundry/blob/main/docs/concepts/patches.md), [custom configuration](https://github.com/SigNoz/foundry/blob/main/docs/concepts/moldings.md#custom-config-files) or [environment variables](https://github.com/SigNoz/foundry/blob/main/docs/reference/casting-file.md#molding-spec) based on your previous configuration.
<details>
<summary><b>Docker Swarm</b> casting.yaml</summary>
3. Review your manifests, we suggest executing the following checks on your manifests before proceeding:
- [ ] Validate the container images match what your deployment had, Foundry uses `latest` on generation by default.
- [ ] If your signoz version was older than latest, please check the [upgrade path](https://signoz.io/docs/operate/upgrade/) first.
- [ ] Check the produced manifests in `pours/deployment` match your older configurations. Extra consideration and review needs to be done on `compose.yaml` as this will be the main entry point for your new deployment.
- [ ] The configuration files for clickhouse are now in YAML so validate your custom settings are present.
```yaml
# casting.yaml
apiVersion: v1alpha1
kind: Installation
metadata:
name: signoz
spec:
deployment:
flavor: swarm
mode: docker
metastore:
kind: sqlite
telemetrykeeper:
kind: zookeeper
telemetrystore:
spec:
config:
data:
config-0-0.yaml: |
macros:
replica: "example01-01-1" # replace with your replica macro
shard: "01" # replace with your shard macro
patches:
- target: "deployment/compose.yaml"
operations:
- op: replace
path: /volumes/signoz-telemetrykeeper-0-data
value:
name: signoz-zookeeper-1
- op: replace
path: /volumes/signoz-telemetrystore-0-0-data
value:
name: signoz-clickhouse
- op: replace
path: /volumes/signoz-metastore-sqlite-0-data
value:
name: signoz-sqlite
- op: add
path: /services/signoz-telemetrykeeper-zookeeper-0/user
value: root
```
4. Execute a `docker compose down`. **Do not** include parameters such as `--volumes` (or `-v`), as it will wipe the volumes we need to maintain and reuse to avoid data loss.
</details>
> [!NOTE]
> This will generate downtime so please plan accordingly.
> The `user: root` patch on the ZooKeeper service lets the container read and
> write the data in your reused ZooKeeper volume, whose files the legacy setup
> created as `root`. Without it, ZooKeeper may fail to start with permission
> errors.
5. Validate the SigNoz containers are down with `docker ps -a`. Multiple containers cannot bind the same volume.
If you had custom configuration (SMTP, extra ingestion receivers/processors,
or custom ClickHouse settings), carry it over via [patches][patches],
[custom config files][custom-config], or [environment variables][env-vars].
6. Run `foundryctl cast -f casting.yaml`. This will recreate the containers based on the spec. This process will download new container images.
### 4. Generate and review the manifests
```bash
foundryctl forge -f casting.yaml
```
Review `pours/deployment/` before deploying:
- [ ] Container images match your current deployment. Foundry generates with
`latest` by default; if your SigNoz version was older than latest, check the
[upgrade path][upgrade-path] first.
- [ ] The generated manifests match your previous configuration, especially
`compose.yaml` (the new entry point for your deployment).
- [ ] The ClickHouse config is now YAML rather than XML; confirm your custom
settings carried over (see [ClickHouse configuration files][ch-config] for
the XML-to-YAML mapping).
### 5. Stop the old deployment
Use the command for your deployment. Do **not** pass `--volumes` / `-v`; that
would delete the data you are migrating.
```bash
docker compose down # Compose
docker stack rm signoz # Swarm
```
> [!NOTE]
> When `cast` is run, the migration container will execute its migrations.
> This causes downtime, so plan accordingly.
## Verifying the Migration
- SigNoz containers will be up and running.
- Log in to the SigNoz UI and verify that data is present.
- Signoz will run on localhost:8080
- Validate that your data ingestion is receiving data.
- Ingesters will receive data on localhost:4317(grpc) and localhost:4318(http)
- Review the logs from both ClickHouse and ZooKeeper; no errors should be present.
Confirm nothing is still bound to the volumes before continuing:
## Rolling Back
Because step 4 brought the legacy stack down *without* `-v`, your original volumes
are untouched and still hold your data. To roll back:
```bash
docker ps -a
```
- Stop and remove the containers created by Foundry (`docker compose down`, again without `-v`).
- Confirm the containers are gone with `docker ps -a` so nothing else is bound to the volumes.
- Reapply your original docker compose file (`docker compose up -d`). It will reattach to the
existing volumes and restore your prior state.
### 6. Deploy with Foundry
```bash
foundryctl cast -f casting.yaml
```
This recreates the containers against your existing volumes and pulls the
images. The migration container runs the schema migrations as part of `cast`.
**Prefer not to use `cast`?** The manifests in `pours/deployment/` are standard
Docker artifacts you can apply yourself. Run the command from that directory so
the relative config paths resolve:
```bash
cd pours/deployment
docker compose up -d # Compose
docker stack deploy -c compose.yaml signoz # Swarm
```
## Verify
- All SigNoz containers are running.
- The UI is reachable on `http://localhost:8080`, and OTLP on `4317` (gRPC)
and `4318` (HTTP), so already-instrumented apps and saved bookmarks keep
working.
- Your existing data is present in the UI, and new data is being ingested.
- ClickHouse and ZooKeeper logs show no errors.
## Roll back
Step 5 left your volumes untouched, so your data is intact. To return to the
previous setup:
1. Bring down the Foundry deployment (`docker compose down` or
`docker stack rm signoz`, again without `-v`).
2. Confirm the containers are gone with `docker ps -a`.
3. Re-apply your backed-up stack: `docker compose up -d` (Compose) or
`docker stack deploy -c docker-compose.yaml signoz` (Swarm). It reattaches
the same volumes and restores your prior state.
## Troubleshooting
- Please reach out to our community on [Slack](https://signoz.io/slack).
If the migration runs into trouble, see
[Troubleshooting Foundry][troubleshooting] for how to capture what we need to
help (the `--debug` output, the exit code, and your `casting.yaml`), then reach
out on [Slack][slack].
## References
- [SigNoz Docker installation docs](https://signoz.io/docs/install/docker/)
- [SigNoz documentation](https://signoz.io/docs)
- [Foundry](https://github.com/SigNoz/foundry)
- [Foundry][foundry]
- [Casting file reference][casting-ref]
- [Custom config files][custom-config]
- [Patches][patches]
- [SigNoz documentation][signoz-docs]
[foundry]: https://github.com/SigNoz/foundry
[install-docs]: https://signoz.io/docs/install/
[compose-example]: https://github.com/SigNoz/foundry/tree/main/docs/examples/docker/compose
[patches]: https://github.com/SigNoz/foundry/blob/main/docs/concepts/patches.md
[custom-config]: https://github.com/SigNoz/foundry/blob/main/docs/concepts/moldings.md#custom-config-files
[env-vars]: https://github.com/SigNoz/foundry/blob/main/docs/reference/casting-file.md#molding-spec
[casting-ref]: https://github.com/SigNoz/foundry/blob/main/docs/reference/casting-file.md
[ch-config]: https://clickhouse.com/docs/operations/configuration-files
[upgrade-path]: https://signoz.io/docs/operate/upgrade/
[troubleshooting]: https://signoz.io/docs/setup/foundry/troubleshooting/faq/
[slack]: https://signoz.io/slack
[signoz-docs]: https://signoz.io/docs