Note: If you're looking for a more hands-off experience we recommend migrating to PostHog Cloud.
Migrating event data using the Replicator App
If you're attempting this migration, feel free to ask questions and provide feedback via the PostHog Communty Slack workspace or a GitHub issue. You should also be aware that some of the steps on this document are potentially destructive! Proceed with caution.
Requirements
- Existing PostHog instance that you can login to (if self-hosting it should be running a minimum version of
1.30.0
) - New PostHog instance
Approach
With this migration, we will migrate over your events using a PostHog app, which will create the necessary person, person distinct ID, and related records in cloud.
We will only be migrating data used to run analytical queries, such as events, persons and person distinct IDs. This means that dashboard configuration, feature flags and other data that is not used in analytical queries will need to be recreated on cloud.
Instructions
Instructions for using PostHog Replicator app:
- Log in to your old PostHog instance
- Click 'Apps' on the left-hand tool bar
- Search for 'Replicator'
- Select the app, press 'Install'
- For the configuration details using the following:
Host
: the hostname of your new PostHog instance e.g.app.posthog.com
if you are using US cloud,eu.posthog.com
if you were are EU cloud, or the domain you'll be sending events to if you are self-hostingProject API Key
: the API key for the new project that you want to send events to - found in the project settings page.
- See events come into the new PostHog instance, identically to the originals
- Once the Replicator is activated, go back to the configuration menu and click the gear next to "Export historical events" under "Jobs." On the calendar, select a start and end date, then click "Run job now." This will begin the process of historical events being import (and might take a while). You can see the progress in logs or by clicking the "App metrics" button (graph) and going to the "Historical exports" tab.
- To track progress of the export you can click the graph icon next to the Replicator app and then click on the "Historical Exports" tab
Migrating non-event data
To migrate the non-event data (such as user information, feature flags, dashboard configurations, etc) we'll need to create a data dump and load that into the new instance. This non-event data is stored in Postgres.
This is only officially supported for importing events into a new cloud instance of PostHog or an enterprise self-hosted instance using Kubernetes. If you're using a different deployment method, you can still try this approach but we can't guarantee it will work.
- You should have a clean enterprise self-hosted PostHog instance up and running. Your new PostHog instance should have no ingested events. We recommend using a fresh and unused installation.
- Your old and new instances should both be running the exact same version of PostHog (minimum
1.30.0
).
Exporting Postgres Data
Please note that this guide assumes your PostHog pods are running in the
posthog
kubernetes namespace, thus appending-n posthog
tokubectl
commands. Substituteposthog
for the namespace applicable to you, or omit the-n
flag if you're using thedefault
namespace.
1. Creating a data dump from your old instance
Access your old PostHog instance and run the following command:
# tip: use the flag -U to specify a username if necessarypg_dump -d posthog -f tmp/export.sql --no-owner --data-only --disable-triggers \-t posthog_action \-t posthog_actionstep \-t posthog_annotation \-t posthog_cohort \-t posthog_dashboard \-t posthog_dashboarditem \-t posthog_featureflag \-t posthog_featureflagoverride \-t posthog_messagingrecord \-t posthog_organization \-t posthog_organizationinvite \-t posthog_organizationmembership \-t posthog_personalapikey \-t posthog_plugin \-t posthog_pluginattachment \-t posthog_pluginconfig \-t posthog_pluginstorage \-t posthog_team \-t posthog_user \-t posthog_user_groups \-t posthog_user_user_permissions
Note: You may see a warning regarding circular dependencies between tables, which shouldn't be an issue as we are using
--disable-triggers
. If you do run into issues when restoring, try doing a full dump instead of--data-only
.
If you do not access your Postgres database via a port-forwarding mechanism to your local machine, you will then need to copy the file from the server to your local system (or the remote server from where you access your new instance). You can do this with scp
:
Note: If your old PostHog instance is hosted on Heroku, you should follow this guide to export all of your data, load into a separate database (we recommend a local Postgres instance), and then run the command above.
scp <your_username>@<database_hostname>:/tmp/export.sql .
If your old instance was deployed using the Helm chart, you can use:
kubectl cp <postgres-pod-name>:/tmp/export.sql ./export.sql # -n posthog
2. Accessing Postgres on your new instance
To access Postgres on your new PostHog instance (Kubernetes cluster), you should do the following:
Tip: Find out your pod names with
kubectl get pods -n posthog
Find out your Postgres password from the web pod:
Terminalkubectl exec -n posthog -it your-posthog-web-pod \-- sh -c 'echo password:$POSTHOG_DB_PASSWORD'Connect to your Postgres pod:
Terminal# Replace posthog-posthog-postgresql-0 with your pod's name if differentkubectl exec -n posthog -it posthog-posthog-postgresql-0 -- shConnect to the
posthog
DB:Terminalpsql -d posthog -U postgresPostgres will ask you for the password. Use the value you found from step 1.
3. Deleting data in the new instance
In order to correctly migrate your data over, we need to make sure the tables we're importing into are completely empty. This will ensure that the foreign key mappings are set up correctly.
To do so, connect to your new Postgres instance using the command from the previous step.
Refresher: Connect to Postgres from your Postgres pod with the command
psql -h your-pgbouncer-host -p 6543 -d posthog -U postgres
.
Once connected, you should have access to the psql
shell. To erase all data in the tables we'll be importing into, run the following:
Warning: Only run this in your new instance.
-- This command is a destructive action that leads to data deletion!!!-- Uncomment the next line (remove '--') if you're certain you want to run it-- TRUNCATE TABLE posthog_action,posthog_actionstep,posthog_annotation,posthog_cohort,posthog_dashboard,posthog_dashboarditem,posthog_featureflag,posthog_featureflagoverride,posthog_messagingrecord,posthog_organization,posthog_organizationinvite,posthog_organizationmembership,posthog_personalapikey,posthog_plugin,posthog_pluginattachment,posthog_pluginconfig,posthog_pluginstorage,posthog_team,posthog_user,posthog_user_groups,posthog_user_user_permissionsCASCADE;
The command output may indicate that some other tables were also truncated. This is normal and expected.
4. Copying your data file to the Postgres pod
In order to copy the data to the new Postgres instance, we must first have the data accessible in our Postgres pod, where we're able to access our instance.
To do so, run the following in the directory of your export (data dump) file:
# Replace posthog-posthog-postgresql-0 with your pod's name if differentkubectl cp export.sql posthog-posthog-postgresql-0:/tmp/export.sql -n posthog
This will copy the export file to a tmp/
directory in the Postgres pod.
5. Loading the data into the new instance
With the export file now in the right place, we can finally import the data!
First, access your Postgres pod:
# Replace posthog-posthog-postgresql-0 with your pod's name if differentkubectl exec -n posthog -it posthog-posthog-postgresql-0 -- sh
Now, run the following to import the data:
psql -d posthog -U postgres < tmp/export.sql
I'm seeing some errors, what should I do?
If you see some errors on this final step, you can try to use pg_restore
instead. You should follow all steps from the beginning, with two modifications:
- When running the
pg_dump
command, add the flag-Fc
. Your command will end up looking something like this:
pg_dump -Fc -d posthog -f export.sql --no-owner --data-only # ...tables
- On this final step, instead of
psql
, usepg_restore
, like so:
pg_restore -U postgres -d posthog < /tmp/export.sql
If errors persist, reach out to us on Slack.
And that's it! If you see no errors, your data should now have been migrated over.
6. Upgrade
As a final step, we should update the cluster with the following command:
helm upgrade -f values.yaml --timeout 30m posthog posthog/posthog --atomic -n posthog
Once the command completes, you can check that all pods are healthy (status = Running
) with the following command:
kubectl get pods -n posthog
The output should look something like this:
NAME READY STATUS RESTARTS AGEchi-posthog-posthog-0-0-0 1/1 Running 0 11dclickhouse-operator-6b5438eh5fb-bt5fk 2/2 Running 0 11dposthog-beat-7782927b778-wxvhl 1/1 Running 0 11dposthog-cert-manager-69fahs7b57-c48dn 1/1 Running 0 11dposthog-cert-manager-cainjector-6d95d93mn8-6tz6k 1/1 Running 0 11dposthog-cert-manager-webhook-6469930mdfc-6l958 1/1 Running 0 11dposthog-events-55283995cc-rpjdm 1/1 Running 0 11dposthog-ingress-nginx-controller-648bdn892f-w7qhp 1/1 Running 0 11dposthog-pgbouncer-77fb4djs85c-2d24t 1/1 Running 0 11dposthog-plugins-54fjd8292649-66gsm 1/1 Running 0 18mposthog-posthog-kafka-0 1/1 Running 0 11dposthog-posthog-postgresql-0 1/1 Running 0 11dposthog-posthog-redis-master-0 1/1 Running 0 11dposthog-posthog-zookeeper-0 1/1 Running 0 11dposthog-posthog-zookeeper-1 1/1 Running 0 11dposthog-posthog-zookeeper-2 1/1 Running 0 11dposthog-web-78dns2f5d7-6zdkc 1/1 Running 0 11dposthog-worker-7857nd8268-j8c4f 1/1 Running 0 11d
Moving a license key
If you have an existing license key, please email sales@posthog.com to move your license key over.