github EasyEngine/easyengine v4.14.0
EasyEngine v4.14.0

one hour ago

Highlights

  • Safer backups and restores. A restore is now a transaction: it checks the free space first, stages everything next to the site, checks the restored nginx config, and rolls back config, files and database if a step fails before the restore is complete (the final container restart comes after that point). Every archive is tested before it's uploaded, a completed backup is never deleted because a later step failed, and retention keeps exactly no-of-backups of this server's backups of each site.
  • Recoverable container upgrades. ee cli update waits for each global service it recreates to be ready (nginx-proxy, the database and Redis must answer a check; cron only has to start) and rolls back to the previous images when one fails. Stopped global services and disabled sites are handled instead of only recording their new image tags.
  • Let's Encrypt fails closed. A site whose certificate couldn't be issued, or is still waiting for manual DNS-01 validation, stays on plain HTTP instead of redirecting to a broken HTTPS.
  • Host logs can't fill the disk. The host's syslog and mail logs get a 2 GB size cap and hourly rotation, and EasyEngine's own logs rotate by rename instead of copying, so a rotation never needs as much free space as the log.
  • Updated images: postfix 3.11.7, OpenResty 1.31.8.1, an nginx-proxy that answers HTTPS for a site without its own certificate with a 503 instead of a 500, and the PHP 8.1–8.5 images (same PHP versions) with safer session defaults.

Upgrade notes

Backups now go to a per-host directory

New backups are stored under <rclone-path>/<site>/<host-id>/<backup-id>. The host id is generated on the first backup or restore and kept in the EasyEngine database; backup-host-id in config.yml overrides it (up to 63 letters, digits, ., _ or -, starting with a letter or digit).

  • Existing backups stay where they are and can still be listed and restored. ee site restore without --id takes this server's newest backup first, then the newest backup in the old layout (with a warning). Backups of other servers are only restored when named with --id=<host-id>/<backup-id>, or with a bare --id=<backup-id> that matches exactly one other server's backup (restored with a warning; several matches are refused).
  • Old-layout backups are no longer pruned automatically, since they may belong to another server using the same rclone-path. Remove them by hand when you no longer need them: list them with rclone lsf --dirs-only <rclone-path>/<site>/, then rclone purge <rclone-path>/<site>/<backup-id> for the backup ids at that level (never the <host-id>/ directories).
  • To keep using the backups of a server this one replaces, set backup-host-id: <old host id> in this server's config.yml, or restore once with --id=<old-host-id>/<backup-id>.
  • Scripts that build <rclone-path>/<site>/<backup-id> paths themselves need to add the host directory. The EasyEngine Dashboard's remote_path already includes it.
  • Older EasyEngine versions don't list or restore backups made by this version, for any site type. Without --id, they restore the newest old-layout backup instead.

What a restore does now

  • A restore replaces all of the site's app/ directory (for WordPress, everything except WordPress core), so files created after the backup are removed. It needs the extracted size of the backup free next to the site, on top of the download, and refuses up front when that space isn't there.
  • A restore refuses to start when something is mounted inside the site directory (for example uploads mounted under app/). Unmount it first; the site is not changed.
  • When a restore replaces the site directory, the site's containers are restarted at the end.
  • A restore that was interrupted is rolled back by the next ee site restore, and ee site backup refuses to back up a half-restored site. After a killed restore, remove the lock file the error names first.
  • Database names used with backup and restore may only contain letters, digits and _ . @ % + = , : -, and can't start with -.

Backup behaviour

  • WordPress backups now hold all of app/ except WordPress core and tool caches, so files such as robots.txt, verification files, .user.ini and custom directories are kept.
  • Symlinks are backed up as links. If a WordPress site's wp-content/uploads points outside the site, only the link is backed up and the backup warns.
  • no-of-backups must be a whole number of at least 1. Any other value falls back to the default of 7 with a warning.
  • When the EasyEngine Dashboard success callback fails, the uploaded backup is now kept and the command fails with error 4004.
  • New error codes: 1004 (the site is half-restored after an interrupted restore; backup refused), 1005 (unsupported database name), 2012 (invalid backup-host-id), 2013 (a required tool is still missing after the install), 3005 (an archive failed its integrity check), 4004 (the EasyEngine Dashboard success callback failed; the backup was kept), 4005 (the database dump couldn't be moved into the backup), 5004 (unexpected error).

Host log rotation and the rsyslog package

On Linux, the update adds maxsize 2G (and a dateformat that allows several rotations a day) to each stanza of /etc/logrotate.d/rsyslog (and /etc/logrotate.d/syslog on older RHEL-style hosts) that has no size limit yet, and makes logrotate run hourly (a logrotate.timer drop-in, or /etc/cron.hourly/ee-logrotate on hosts without the timer). The original rule is saved in /opt/easyengine/.backup/logrotate/. This happens once: if you change or remove the cap later, updates don't put it back.

/etc/logrotate.d/rsyslog is a dpkg conffile. Because EasyEngine edits it, a later rsyslog package update that changes this file is held back by unattended-upgrades ("conffile prompt … needs to be upgraded manually"), and apt upgrade asks which version to keep. Upgrade rsyslog by hand (sudo apt-get install --only-upgrade rsyslog) and either keep the installed version or take the package's version and add maxsize 2G back to each stanza.

EasyEngine's own log rules (ee_cli, ee_sites, ee_nginx_proxy) get maxsize 1G and now rotate by renaming the log and having nginx and php-fpm reopen it, instead of copytruncate.

Containers are updated

  • The update moves the global services and enabled sites to the new images listed below. Only the global services it recreates are waited for: nginx-proxy, the database and Redis must answer a readiness check, and cron only has to start. If one isn't ready in time, the update rolls back to the previous images.
  • A stopped nginx-proxy, Redis or New Relic daemon keeps its old image until you enable it, and ee service enable then recreates it on the new image. Missing global services are recreated.
  • Disabled sites get updated compose files and pick up the new images when you enable them.
  • Sites whose Let's Encrypt certificate is still pending stay on HTTP during the image migration.
  • Each recreated global service gets up to 60 s to become ready, and the database 300 s. The new global-service-ready-timeout key in config.yml (seconds) replaces the wait for every service, the database included, so set it above 300 (for example 600) if a slow database start makes an update roll back.
  • docker-compose is installed (2.27.0) only when it is missing or older than 2.26.1, and never downgraded.
  • On PHP 8.4, 8.5 and latest sites, the update removes the deprecated session.sid_length and session.sid_bits_per_character lines from the site's php.ini and saves the previous file as /opt/easyengine/.backup/session-sid-settings-<YYYYMMDD-HHMMSS>/<site>/php.ini. It warns only when a removed line held a value you set yourself; the lines shipped by EasyEngine are removed silently.

PHP requirement checks

ee now checks the PHP extensions and functions it needs before a command runs, and names the disabled functions and the ini files that disable them (for example on HestiaCP servers). A failed requirement check now exits with status 1, but ee cli update still runs (with a warning) when other PHP requirements are missing, and ee cli info shows a "PHP requirements" row. Site commands keep working when pcntl_* or shell_exec are disabled.

Fixes and improvements

  • Let's Encrypt (EasyEngine/site-command#521, EasyEngine/site-type-wp#247, #1955): a failed Let's Encrypt create leaves the site on plain HTTP with an ee site update <site> --ssl=le hint. Cloudflare DNS-01 waits for the zone's authoritative nameservers (up to 120 s) instead of a fixed 10 s, a manual DNS-01 site stays on HTTP until ee site ssl-verify, and WordPress URLs follow the scheme the site is actually served over (for subdomain multisite, the subsites' URLs too; a failed switch prints the search-replace command to run). The www redirect is kept when only the HTTPS hop of the www check fails, and ee site ssl-renew on a pending site points to ee site ssl-verify.
  • Backup and restore (EasyEngine/site-command#527, #483, #481): restore transaction and rollback, archive integrity checks with 7z t, database credentials passed through option files instead of the command line, the database dump kept out of the webroot, missing tools installed without prompts, unique backup ids, and a consistent EasyEngine Dashboard callback flow (at most one final callback, the global backup lock released before the callback retries).
  • Upgrades (#1953, #1954, EasyEngine/service-command#84): service readiness checks with rollback, a retry after an interrupted upgrade keeps the rollback snapshot, no needless image pulls or docker-compose downloads, and stopped services and disabled sites handled.
  • PHP session settings (EasyEngine/site-command#519, #520, EasyEngine/dockerfiles#372, #374): PHP 8.4 and 8.5 no longer carry the deprecated session.sid_* settings, and session.upload_progress.enabled stays Off on PHP 8.1–8.5 in the images, in ee site update --php and after restoring an older backup. The update edits the php.ini of existing PHP 8.4, 8.5 and latest sites and keeps a copy of the previous file (see "Containers are updated").
  • Site config volumes (EasyEngine/site-command#522, EasyEngine/site-type-wp#248, EasyEngine/site-type-php#121): PHP config files in a site's config volume are read and written only as regular files (symlinks are refused), and ee site update --php pulls the new image first and rolls back on failure (EasyEngine/site-command#522).
  • PHP requirements (#1957, EasyEngine/site-command#523): clear errors for disabled PHP functions; site commands work without pcntl_* or shell_exec.
  • Logs (#1958, #1960): host syslog and mail log size cap with hourly rotation, and rename-and-reopen rotation for EasyEngine's own logs.
  • phpRedisAdmin (EasyEngine/admin-tools-command#56): /ee-admin/pra/ works again behind the admin-tools authentication with current phpRedisAdmin releases (1.27.0 returned 403). There is no migration: on a site that already has admin-tools, run ee admin-tools enable <site> --force once.
  • EasyEngine Dashboard naming (EasyEngine/site-command#525, EasyEngine/dash-command#3): messages and help say "EasyEngine Dashboard". Flags and config keys are unchanged.

Docker image updates

Image Before After Change
easyengine/nginx-proxy v4.13.0 v4.14.0 Debian updates (EasyEngine/dockerfiles#369); an HTTPS request for a site without its own certificate (including HTTP-only sites) gets the default certificate and a 503 instead of a 500; on a server missing the default certificate the TLS handshake for such a site is now rejected instead of being answered with another site's certificate (EasyEngine/dockerfiles#300)
easyengine/nginx v4.11.0 v4.14.0 OpenResty 1.31.8.1 (EasyEngine/dockerfiles#377)
easyengine/postfix v4.13.1 v4.14.0 postfix 3.11.7 with /etc/aliases (EasyEngine/dockerfiles#368)
easyengine/php8.1 v4.13.0 v4.14.0 session.upload_progress.enabled Off (EasyEngine/dockerfiles#374)
easyengine/php8.2 v4.13.3 v4.14.0 session.upload_progress.enabled Off (EasyEngine/dockerfiles#374)
easyengine/php8.3 v4.13.3 v4.14.0 session.upload_progress.enabled Off (EasyEngine/dockerfiles#374)
easyengine/php8.4 v4.13.3 v4.14.0 deprecated session.sid_* settings dropped, upload progress Off (EasyEngine/dockerfiles#372, #374)
easyengine/php8.5 v4.13.3 v4.14.0 deprecated session.sid_* settings dropped, upload progress Off (EasyEngine/dockerfiles#372, #374)
easyengine/redis v4.13.1 v4.14.0 rebuilt, no Dockerfile change
easyengine/cron v4.11.0 v4.14.0 Alpine updates (EasyEngine/dockerfiles#370)
easyengine/newrelic-daemon v4.13.1 v4.14.0 Alpine 3.24.3, daemon 12.12.0.41 (EasyEngine/dockerfiles#371)

All other images (mariadb, mailhog, PHP 5.6–8.0 and easyengine/php) are unchanged.

Known issues

  • ee cli self-uninstall leaves the edited /etc/logrotate.d/rsyslog, the hourly logrotate drop-in or cron job, and EasyEngine's own logrotate rules in place, and removes /opt/easyengine/.backup/logrotate/ with the copy of the original rsyslog rule. Copy that backup elsewhere first if you plan to uninstall.

What's Changed

Don't miss a new easyengine release

NewReleases is sending notifications on new releases.