Migration

Updated Nov 18, 2026 Markdown

Imports accounts over SSH from cPanel or DirectAdmin, the two sources measured end to end. The wizard also detects Plesk, HestiaCP, CyberPanel, CloudPanel and CWP on the source server, and it can import from another WHost server through that server's API (no SSH); those sources have not been measured yet and migration is experimental in the beta (see Known Limitations). An import from another WHost server writes the account's public_html (replaced) and mail (merged) through directory handles as the account's user — also when the account already exists here: a link the account placed at public_html is replaced by a plain folder, a link at mail makes that account's import fail rather than being written through. The panel follows each job by polling its status; the API also offers the progress as an SSE stream (GET /migration/status/{job_id}/stream).

1Source server credentials (SSH host, port, root/sudo user, password or key)
2Scan — agent pulls the account list and lets the operator pick which to migrate (filter by domain pattern). The scan runs in the background: the wizard and the History tab count the scanned accounts against the accounts found (a large server can take several minutes). Leaving the page does not stop the scan and the History tab keeps counting. A finished scan's History row ends with an action: Select Accounts opens that scan's account selection while its SSH credentials are still in memory (30 minutes after the scan; an agent restart drops them); after that the action is Scan Again, which opens the connection form filled in with the source's host, port, SSH user and panel — enter the password or key and scan again. Starting a migration whose credentials are gone leads to the same form. A scan that cannot connect, cannot detect the panel or stops on a command the source server does not finish (30 seconds per command, 300 seconds for an account's disk usage, where an account that does not finish is listed with 0 MB) ends as Failed with the reason. The scan takes what the source panel already records instead of walking each home: on cPanel every account's disk usage from cPanel's own account list (whmapi1 listaccts, the figure WHM shows, which also counts mail and other files the account owns) and every account's database count from one query; on HestiaCP each user's disk usage and database count from the user list it reads once (v-list-users); on Plesk every subscription — system user, status, service plan, PHP handler, disk usage (Plesk's own statistics) and database count — from one query on Plesk's database (plesk db), the mailboxes per subscription; on DirectAdmin, CWP and CyberPanel every account's database count from one query. An account the panel reports no usage for (a server without disk quotas, a new or empty account) is measured with du, and so is every account on DirectAdmin, CWP, CyberPanel and CloudPanel, which keep no usage figure that can be read
3Start — per account, in this order: create the account here, copy the files (.htaccess rules checked, LSCache detected; a CyberPanel account is its website, whose files come from that website's home /home/<domain>/public_html), the databases and the mailboxes (each can be skipped), then cron jobs and FTP accounts. SSL certificates are not carried over — issue them again here (see the FAQ). The account is created with the e-mail address the source panel records for it (cPanel, DirectAdmin, HestiaCP, CWP, another WHost server), otherwise with the administrator's e-mail (Settings → Profile); with neither it fails with that reason. An account whose files or databases were copied only partly — the source server's archive or database dump ended with an error, an archive did not unpack, a database did not import — is marked failed with the reason (the folder or the database name); files that changed on the source while they were read are accepted
4Follow the job until every account is finished or failed
5Cut-over — check the sites and mail here, then point the domains' DNS at this server; no DNS records are exported or compared for you

Failed accounts can be retried while the scan's SSH credentials are still in memory; accounts that finished are not run again. An agent restart does not resume a running job: it is marked failed ("Server restarted during migration") and the credentials are gone, so scan the source server again before retrying (400 MIGRATION_CREDENTIALS_EXPIRED otherwise). SSH credentials are AES-encrypted in memory and zeroed after the job completes.

What a migrated site gets here. The account's databases and database users on the source (<username>_*) come over; from a cPanel source also every database cPanel's own record (WHM API 1) gives the account and the users of those databases — cPanel long named a database with the first 8 characters of a longer account name, and such a database keeps that short prefix. Plesk and CyberPanel do not start database names with the account name, so from them the panel's own record alone decides which databases and users are the account's: Plesk's (its psa database, read with plesk db) gives the subscription's — its main domain's and its other domains' — and CyberPanel's the website's; a Plesk database kept on another database server is named in the warnings. CloudPanel keeps no record of a site's databases that can be read: from it only <username>_* comes over and every CloudPanel account's warnings say so — move another database with clpctl db:export on the source and an import on the Databases page. The source's database server is read with the login its panel keeps for it: DirectAdmin's conf/my.cnf, Plesk's plesk db, the root password in CyberPanel's /etc/cyberpanel/mysqlPassword (read on the source, never sent here), root's own login on cPanel, HestiaCP, CWP and CloudPanel. A database server that refuses that login and a database whose name this server cannot take are named in the warnings. Here an account's databases and their users are named <username>_…, so one with another prefix (or none) is created as <username>_<rest> (abcdefgh_wp of the account abcdefghij becomes abcdefghij_wp); a new name that is already taken here (the source's own <username>_… names go first) or longer than this server allows (64 characters for a database, 32 for a user) is not created and is named in the job's warnings. The site's wp-config.php — in public_html, in a folder right below it or in the home — gets the new DB_NAME / DB_USER where its value is exactly the old name; every rename and every changed file is a warning of the account, and an application other than WordPress keeps the old names in its own configuration until you change them. The users come over with the source's password hash and every privilege on the databases that came over, so the site's own configuration connects as before; a password never travels in clear. Only a mysql_native_password hash can be taken over: a user stored with another method (MySQL 8's caching_sha2_password), a name outside the account's prefix that the panel's record does not give it, or one this server does not accept is not created and is named in the job's warnings — create it on the Databases page with the password the site uses. The migration page shows each account's warnings: on the account's row while the job runs and when it is complete, and under History the job's with warnings count opens them (GET /migration/status/{job_id}, accounts[].warnings; the history list counts them in warned_accounts). The scan's database count comes from cPanel's record too (PostgreSQL databases are not counted). A user that already exists here keeps its password and gets the grants. After the copy, the home and public_html go back to the mode and group the account's web server reads them with (0750 and the web group on every web server), and the copied files keep the modes they had on the source. The default page the account was created with (public_html/index.html) is removed when the copied site brings an index.php or index.htm — Apache would serve it in front of the site — and stays readable otherwise; an index.html the site brought is left alone. The account takes the source's PHP version when that version's PHP-FPM runs here, otherwise the server's default (the agent log says so). From DirectAdmin that is the version of the CustomBuild slot the domain selects (php1_select=N in the domain's .conf picks phpN_release in CustomBuild's options.conf; a domain without a selection runs the first slot); from HestiaCP the version in the name of the web domain's PHP-FPM backend template (PHP-8_2), while a domain on the default template takes this server's default. At the end of the job the WAF's application exclusion sets are brought to the copied sites at once (a site WordPress is found on gets the WordPress set), not at the agent's next round within five minutes.

Boundaries the transfer keeps. The source server's SSH host key is pinned on first use (/var/lib/whost/migration_known_hosts); a later mismatch aborts the connection with the fingerprint in the message. Files and mailboxes are unpacked on this server as the migrated account's own user, so nothing the source archive contains can reach outside what that account may already write. Databases are created locally only under the account's own prefix (<username>_…): a source name outside it is created under the prefix when the source panel's record gives the database to the account (above), and not at all otherwise. Their dumps are imported the way a backup restore imports them — by the account's import user, which can write to the account's own databases only (see Backups — scope and settings): a dump that names another database is refused at that statement and the agent log names the database. The private/loopback/link-local address ranges are refused as a source host. In the new account's home the transfer works through directory handles: the extraction folder (public_html, mail) must be a plain folder ("Refusing to extract into … not a plain directory" otherwise), and the ownership fix after the file transfer and the SSL certificate transfer leave a link the account placed at public_html or ssl alone and name it in the agent log ("ownership of … not fixed", "SSL certificates not transferred …") instead of following it; certificate files are written as new root 0600 files. The .htaccess conversion reads the file the same way: a .htaccess that leads out of the account's home is not read and the migration warnings say "Could not read …".

Restoring from an uploaded backup file is not available. Migrate over SSH with the steps above. The panel has no Upload tab, and POST /migration/upload and POST /migration/upload/restore answer 409 MIGRATION_UPLOAD_RESTORE_UNAVAILABLE without keeping the file.

API endpoints: POST /migration/scan, GET /migration/scan/{job_id}, POST /migration/start, GET /migration/status/{job_id}, GET /migration/status/{job_id}/stream (SSE), POST /migration/cancel/{job_id}, POST /migration/retry/{job_id}, GET /migration/history, GET /migration/history/{job_id}.

Still Need Help?

Our support team is here around the clock for anything you can't find above.