Tüm Sürümler
1.0.0-beta.3
This release brings 3 new features, 3 security fixes and 24 fixes.
Yeni Özellikler
- The web application firewall now protects sites served by OpenLiteSpeed. On a server whose web server is OpenLiteSpeed, ModSecurity with the OWASP Core Rule Set and the WHost rules now checks and blocks the requests to every site's applications (PHP pages and other dynamic content), as it does under Nginx: each account's WAF switch, mode and rule exclusions apply, blocked requests are listed on the WAF pages and the WAF page shows the firewall as active. A request for a static file (an image, a stylesheet, a script, an HTML file) is not checked once OpenLiteSpeed has served that file; the WAF page says so and points to the firewall for blocking an address. The module is installed when OpenLiteSpeed is switched on, and on servers already running OpenLiteSpeed shortly after the update.
- Custom WAF rules can be added, edited and removed from the admin panel. Administrators can keep their own ModSecurity rules under WAF > Custom rules: each rule is checked before it is saved (only rule directives, no operator that reads files or asks an outside service, no exec action, a rule id outside the ranges the Core Rule Set and WHost use, the rule engine's parser and the web server's configuration test), a refused rule is shown with the line and column at fault, and accepted rules apply to every site under Nginx, and under OpenLiteSpeed to application requests (there a static file is not checked once it has been served). Rules written by hand in the custom rules folder keep working and are listed read-only.
- LiteSpeed Enterprise: WAF through LiteSpeed's own engine (off until switched on). With modsecurity.litespeed_enterprise set to true in /etc/whost/agent.conf, the WAF rule set (OWASP CRS, WHost rules, custom rules, per-account switch, mode and exclusions) is applied on LiteSpeed Enterprise: a server-level rule set named WHost with the engine off there, and each site whose account WAF is on switches it on. The WAF request lists (the WAF page's Audit Log, the account's and a reseller's sub-account WAF panel) and the WAF alerts include LiteSpeed's rule matches, read from its error log because its engine writes no audit log: such an entry reads Matched, since LiteSpeed does not record whether the request was refused (with the account's mode Block it was), and carries no request body, headers or response. The setting is off by default; setting it back to false removes the rule set and the sites' switch within five minutes.
Güvenlik
- Only root and the web server can connect to the LiteSpeed listener. On a server running OpenLiteSpeed or LiteSpeed Enterprise, only root and the web server (Nginx in front of LiteSpeed) can open a connection to the LiteSpeed listener behind Nginx (port 8088), on any of the server's own addresses; processes of hosting accounts are refused. A local process can no longer name the visitor address the sites' access logs start with, so it can no longer have the web jails ban an address of its choosing. Sites reached by their names over ports 80 and 443 are not affected; a script that connected straight to port 8088 now gets 'connection refused'. The rule is set when OpenLiteSpeed or LiteSpeed Enterprise is activated and at every agent start, and removed when the server goes back to Nginx or Nginx + Apache.
- Webmail runs Roundcube 1.6.21; existing servers are upgraded at the agent's start. Webmail moves from Roundcube 1.6.9 to 1.6.21, which closes the security issues fixed in between. New installations get it at once; an existing server is upgraded in the background a few minutes after the update, keeping its configuration and the previous release beside it.
- Webmail and phpMyAdmin no longer serve their internal files. The web server now runs only the entry points of webmail (Roundcube) and phpMyAdmin and refuses their library, plugin, example, documentation and package files, also on servers in the OpenLiteSpeed or LiteSpeed Enterprise mode. phpMyAdmin's help icons open the same page on docs.phpmyadmin.net.
Hata Düzeltmeleri
- The installer stops at once on a server that is not 64-bit x86. On an ARM64 (aarch64) or any other non-x86_64 server the installer now stops when it starts, before it installs anything, and says that WHost needs a 64-bit x86 (x86_64 / amd64) server; before, it installed the whole stack and the agent then failed to start. The installation guide, the supported operating systems page, the FAQ, the getting-started page and the known limitations state the requirement.
- A Let's Encrypt certificate that is not issued now shows why. When Let's Encrypt does not issue a certificate, the panel now shows the reason it gave for each name - for example that the www. name has no DNS record, or that another server or a proxy answered the check - in the error message and under the Let's Encrypt form of the install window (admin and client panels), on each failed line of Secure All Domains and on the server name certificate card. Before, the message only said that the request had failed. The same reason reaches the agent log and the error records a bug report can attach, with names and addresses masked.
- Switching to OpenLiteSpeed works again after it was switched off once. After OpenLiteSpeed had been switched off on a server, switching it on again under Plugins failed with "Unit lsws.service not found" and the server stayed on its previous web server. The agent now starts, stops and reloads OpenLiteSpeed by its own service name, so the switch works however often it is made, and the HTTP ban list reaches OpenLiteSpeed while it serves the sites.
- OpenLiteSpeed shows as running on the Services page; restarts and renewals reach it. On a server where OpenLiteSpeed had been switched off once, the Services page showed the web server as stopped while it was serving sites, its Restart and Reload buttons (and Restart LiteSpeed on the Plugins page) failed with "Unit lsws.service not found", and a renewed Let's Encrypt certificate was only served after the next restart of the web server. They now act on the server's own unit (lshttpd); the same holds for LiteSpeed Enterprise.
- Spam filtering runs on AlmaLinux, Rocky Linux and CentOS Stream 9 again. Rspamd's own package for the RHEL family (4.3.0, published on 13 November 2026) installs without the system account the service runs as, so on a new AlmaLinux, Rocky Linux or CentOS Stream 9 server with mail, Rspamd could not start and the installer stopped at its final check without showing the panel address and the admin password. The installer now creates the account when the package did not and hands Rspamd's state and log folders and its Redis settings to it. A server that was already installed this way is repaired when the agent starts after the update: the account is created, the folders and settings are handed to it and Rspamd is started; the agent log says what it repaired. Ubuntu servers were not affected.
- OpenLiteSpeed keeps working when restarts and reloads come in quick succession. On servers running OpenLiteSpeed or LiteSpeed Enterprise, two or three actions on the web server within a few seconds (a restart on the Services page, account, certificate or WAF changes, HTTP-layer bans) could leave the web server in a restart loop in which configuration changes were no longer applied until it was fixed by hand. Actions on the LiteSpeed server now run one at a time, each after the previous graceful restart has finished; a restart pressed right after another one answers a few seconds later.
- Switching to OpenLiteSpeed works on Ubuntu and Debian servers. On Ubuntu and Debian servers, switching to OpenLiteSpeed under Plugins failed after about ten seconds with "Unable to locate package openlitespeed" and the server stayed on its previous web server; every retry failed the same way. The LiteSpeed repository files came out readable only by root, so the package manager ignored the repository. The agent now makes those files readable and reads the package list again, so the switch completes, also on servers where an earlier attempt failed.
- The installer shows the admin password even when a final check fails. When every step succeeded but one of the installer's final checks failed (for example a service that did not start), the installer stopped before its closing screen, so the panel address, the admin password and the API key and secret were never shown and the admin password had to be reset by hand. The closing screen is now always shown: it says that some checks failed and points to the failed lines above it, and the installer still ends with an error so automated installs notice it. Installs where every check passes look the same as before.
- Switching to LiteSpeed Web Server works again (trial and licence). Activating LiteSpeed Web Server under Plugins, with a trial or with a serial key, failed after about ten seconds with "Permission denied" and the server stayed on its previous web server. The downloaded LiteSpeed package lost its program permissions when it was unpacked, so its licence check could not run. The package now keeps them and the activation goes through.
- Switching from Nginx or Apache to LiteSpeed Web Server moves the sites. A switch from Nginx or Apache to LiteSpeed Web Server reported success while none of the sites had been moved into LiteSpeed: every site then answered LiteSpeed's 'Page Not Found'. The server is now installed first and the sites are moved into its configuration after it; if a site cannot be moved, the switch fails, every site goes back to the web server it came from and the activation dialog names the site and the reason.
- OpenLiteSpeed and LiteSpeed Enterprise serve 0640 files and write the sites' access logs. On a server running OpenLiteSpeed or LiteSpeed Enterprise, the default page of a new account and files uploaded in the File Manager no longer answer 403, and every visit reaches the account's access log again with the visitor's address: the Logs pages, the bandwidth counter and the web jails that ban scanners see them (on LiteSpeed Enterprise a request for a hidden file such as /.env that the server refuses on its own still shows 127.0.0.1 and does not count toward a ban). The LiteSpeed server now runs with the web group, as Nginx and Apache read these files; PHP still runs as the account. On OpenLiteSpeed a name no site serves gets a 404 instead of the package's example site. Existing servers are fixed at the first start after the update (one graceful restart). AlmaLinux, Rocky Linux and CentOS Stream, and clean installs on all five supported systems, are measured in the release's clean-install round.
- LiteSpeed Web Server keeps serving the sites when the agent restarts. On a server switched to LiteSpeed Web Server (Enterprise), the web server now runs as its own system service: restarting, redeploying or updating the WHost agent no longer stops it, so the sites keep answering instead of returning 502 until LiteSpeed is restarted by hand. The activation and the LiteSpeed update stop the server the installer started and start it once more through its service (a few seconds, inside the switch). On a server already switched, the first agent start after the update starts LiteSpeed once through its service. A server tuning change restarts LiteSpeed through its service as well. The OpenLiteSpeed update action restarts OpenLiteSpeed through its service too.
- Updating OpenLiteSpeed through the API updates its package. The OpenLiteSpeed update (POST /api/v1/plugins/openlitespeed/update) used to answer 'Plugin updated successfully' without changing anything. It now upgrades the openlitespeed package from the LiteSpeed repository with the server's package manager (apt on Debian / Ubuntu, dnf on AlmaLinux / Rocky / CentOS Stream) and restarts OpenLiteSpeed through its service when the version changed. When the package already is the newest version, the answer says so ('Plugin is already the newest version') and nothing is restarted; the answer names the package versions before and after (data.update_result).
- The panel and the sites stay up after a reboot on LiteSpeed servers. On a server switched to OpenLiteSpeed or LiteSpeed Web Server (Enterprise), installing an SSL certificate no longer adds a listener on port 443 to the LiteSpeed configuration. nginx serves HTTPS in front of LiteSpeed; that listener took port 443 whenever LiteSpeed started before nginx (a boot starts the two side by side), nginx then failed to start and the panel and every site were unreachable. A listener on port 443 already in the LiteSpeed configuration is removed when LiteSpeed is activated and at the first agent start after the update, after LiteSpeed's configuration test; LiteSpeed then reloads its configuration once, without interruption. If LiteSpeed already holds port 443 when the agent starts (nginx could not start at an earlier boot), the agent stops LiteSpeed completely, starts it again through its service and starts nginx, so the panel and the sites come back.
- Sites on OpenLiteSpeed keep answering after an SSL certificate is installed or removed. On a server switched to OpenLiteSpeed, installing or removing a site's SSL certificate (from either panel or the API, and the switch for a name with a certificate) took the site off the listener nginx passes it to, and the site answered 404 over HTTP and HTTPS until its configuration was written again. Installing and removing a certificate now leaves the site where it is. At the first agent start after the update, a hosted site that lost its place this way gets it back, after OpenLiteSpeed's configuration test (one reload, without interruption).
- OpenLiteSpeed / LiteSpeed Enterprise sites served over HTTPS with their own certificate. On a server running OpenLiteSpeed or LiteSpeed Enterprise every site's HTTPS was served with the panel's certificate, so a browser showed a certificate warning for any site without a CDN in front. nginx, which holds ports 80 and 443 in front of the LiteSpeed server, now ends each site's HTTPS with the site's own certificate (Let's Encrypt or uploaded) and passes the request to the LiteSpeed server as before; static files, the WAF, the site's logs, redirects and suspensions stay with the LiteSpeed server. At the first agent start after the update every hosted site gets its nginx file (one render, configuration test, one reload without interruption). These sites' HTTPS is served over HTTP/2.
- Saving the server name, IP or force-SSL keeps OpenLiteSpeed / LiteSpeed sites online. On a server running OpenLiteSpeed or LiteSpeed Enterprise, saving the hostname, the server IP or the force-SSL setting rewrote the panel's nginx file and dropped its forwarding to the LiteSpeed server: every site stopped answering until the web server was switched again. The rewrite now keeps the forwarding, and the first agent start after the update puts it back where an earlier save removed it (configuration test first).
- Uploads over 1 MB work on OpenLiteSpeed and LiteSpeed Enterprise sites. nginx in front of the LiteSpeed server refused request bodies over 1 MB (413), so larger uploads (media, form attachments) failed. The limit is now 256 MB, as on nginx and Apache sites; the LiteSpeed server's own limit and the account's PHP limits still apply.
- A web server switch that cannot move every site is reported and undone. Switching to or from OpenLiteSpeed / LiteSpeed Enterprise removed an account's old site files with the server-wide web server instead of the account's own, so an nginx + Apache account on an nginx server kept its Apache file; once that account was deleted Apache's configuration test failed, and switching back left every site without a configuration while reporting success. The account's own files are now removed; a switch first runs the target's configuration tests and is refused, with nothing moved, when one already fails; and a switch that cannot move every site moves them back, leaves the web server as it was and names the sites in its error. Accounts move one at a time.
- Switching LiteSpeed off now waits until every LiteSpeed process has ended. Switching OpenLiteSpeed or LiteSpeed Enterprise off returned while LiteSpeed processes were still running: a process still finishing a request kept listening on LiteSpeed's local port for as long as that request lasted, and for a moment after the switch-off LiteSpeed still answered local requests that its port guard no longer held back. The switch-off now waits until LiteSpeed has fully stopped - processes that do not end within about half a minute are ended - and lifts the port guard only once nothing listens on that port. When leaving LiteSpeed Enterprise for nginx or nginx + Apache, the sites move to the previous web server before LiteSpeed stops, so they keep answering during the wait. The switch-off takes a few seconds longer; a request still running in LiteSpeed at that moment is ended after about half a minute.
- The Restart PHP Processes button of the LiteSpeed plugins now restarts PHP. In the Manage window of an active OpenLiteSpeed or LiteSpeed Enterprise plugin, Restart PHP Processes was refused every time (Restart failed: the service is not in the monitored list) and PHP kept running as before. It now restarts LiteSpeed gracefully, which renews every account's PHP processes while the sites keep being served.
- The LiteSpeed plugin cards show the installed server version. The OpenLiteSpeed card's version badge showed v1.8 although OpenLiteSpeed 1.9.4 was installed (the Manage window already showed 1.9.4); the LiteSpeed Enterprise card showed v6.3 for 6.3.7. An active web server plugin's card now shows the installed version; an inactive one keeps the plugin's own version.
- Switching to OpenLiteSpeed from LiteSpeed Enterprise waits until it has stopped. Enabling OpenLiteSpeed while LiteSpeed Enterprise was active stopped LiteSpeed Enterprise without waiting: a LiteSpeed Enterprise process still finishing a request kept running next to the newly started server after the switch had reported success. The switch now stops LiteSpeed Enterprise completely before it continues: a process still running five seconds after the stop is ended, so a request running at that moment is cut instead of keeping an old server alive.
- Enabling OpenLiteSpeed while LiteSpeed Enterprise is active switches to OpenLiteSpeed. Enabling OpenLiteSpeed while LiteSpeed Enterprise was active reported success but started LiteSpeed Enterprise again with OpenLiteSpeed's configuration on a server where OpenLiteSpeed had been used before: the sites fell to LiteSpeed's example site, both web server plugins showed as inactive and the panel offered no way back. The switch now puts OpenLiteSpeed back as switching LiteSpeed Enterprise off does: it reinstalls the OpenLiteSpeed package and its ModSecurity module over the LiteSpeed Enterprise files and restores OpenLiteSpeed's service before the sites move. The package reinstall makes the switch about half a minute longer, during which the sites do not answer.
Bu Sürümde
- Yayın
- Yeniden Başlatma
- Güncelleme sonrası gerekli