Release notes.
What changed in JSOPanel, one version at a time. Newest first.
v0.1.265
13 October 2026
- Changed: on the WordPress toolkit (Applications > WordPress), each site now starts collapsed - showing just the name, version and how many plugins and themes need updating - with a Manage button to open everything else for that one site. With several WordPress sites on an account, the page used to show every one of them fully expanded at once; now only one is open at a time.
v0.1.264
13 October 2026
- Fixed: a domain switched to deliver its mail through Microsoft 365, Google Workspace or another outside provider could silently revert to delivering into mailboxes on this server again, with nothing anywhere having been touched - the server's own nightly repair work was partly to blame, and would put it back the next time it ran for any reason. Mail would then be delivered in two different places at once until somebody noticed and switched the setting back by hand. The repair no longer does this; a domain stays wherever it was deliberately set until someone changes it.
v0.1.263
12 October 2026
- Changed: the "Admin password" button on an installed application's row is now labelled "Change admin password" - it was easy to mistake for a way to look up the current password rather than set a new one.
v0.1.262
12 October 2026
- Fixed: an extra IP address added through the panel - IPv6 in particular, when it had its own gateway - could come back unreachable after every reboot, even though the address itself was still there. The only fix until now was restarting a service by hand. The underlying ordering mistake is corrected, and the server now checks every couple of minutes that its own addresses are actually working and quietly puts anything back that isn't, so this can no longer require a manual restart again.
v0.1.259
12 October 2026
- Fixed: the nightly update could get stuck indefinitely if the server had a temporary network problem while fetching software updates - with no way to recover except waiting it out, sometimes for a very long time. Several places where this could happen, including one checked for every website running on the OpenLiteSpeed web server, now give up cleanly after a short wait and try again automatically on the next scheduled update instead of hanging.
v0.1.258
12 October 2026
- Fixed: a visitor loading a WordPress site's login page, or a plugin making ordinary XML-RPC requests, could be refused for no reason connected to that site at all. A handful of shared security rules counted this across every WordPress site on the server rather than one at a time, so a real problem on one site could block an unrelated visitor on a completely different one. Six such rules are now switched off by default on every server; the detection behind them keeps logging genuine attacks, it just no longer blocks on their say alone.
v0.1.257
10 October 2026
- New: temporary allow or block for one address, under Security Center, then Firewall, in the new Temporary allow / block card. Choose allow or block, type the address (or a range), and say how long, in minutes, hours or days, up to 90 days. The rule removes itself afterwards, within about a minute of the time, so nothing has to be remembered. The card lists every temporary rule with when it ends and a Remove now button, and adding the same address again just changes how long it lasts.
- New, for administrators: Domains, a domain's settings, PHP tab now has a Restart PHP button, so nobody has to remember the command. If the account runs its own PHP, only that account's PHP is restarted. If PHP is shared with other accounts, it is reloaded instead - running requests finish and new PHP workers take over - because a restart would cut every other account's requests. The settings are tested first, so a broken PHP setting leaves everything running as it was.
- A temporary block is refused if it would shut out the address you are using, this server's own address, or a very wide range, so a mistake cannot lock you out of the panel.
v0.1.256
6 October 2026
- New, off by default: forgery protection for incoming mail, under Service Configuration, then Spam filter. When switched on, every incoming message is stamped with what this server found out about the sender, and any such stamp the sender typed in themselves is removed first. A program that relies on that stamp, such as a ticket system, can then trust it. Mail sent by a script on the same server is covered too.
v0.1.255
6 October 2026
- Accounts, Users & Resellers, Tools, Move accounts to a different owner now has a search box. Type part of an account name to narrow the list; accounts you already picked stay picked while you search.
v0.1.253
5 October 2026
- New, off by default and for administrators only: remote database access. A customer who needs to connect to their database from their own computer, or from another service, can now be allowed in one address at a time. It works for MariaDB, PostgreSQL and MongoDB, under Databases, then Remote access.
- Each remote connection gets its own login and password, separate from the website's own, and that login can reach one database only. MariaDB and PostgreSQL also limit it to 20 connections at a time. Only a single IPv4 address can be allowed, never a range, and only for databases created through the panel.
- Switching on an engine, allowing an address and removing it each set up the database engine and the firewall together, so they cannot disagree. If a step fails, the earlier steps are undone and you are told why. It will not switch anything on while the firewall is off.
- Connections are encrypted wherever the engine can do it: MariaDB and PostgreSQL logins refuse a connection that is not. If an engine has no encrypted connections set up, the page refuses unless you say you accept an unencrypted one, and the list marks that address. MongoDB is refused if it does not already require a login.
- Removing an address ends its open connections at once. Deleting a database or an account removes its remote access automatically, and uninstalling the panel closes everything.
v0.1.252
4 October 2026
- Backups can no longer replace a good copy with a bad one. A copy that is missing a database, or whose website archive could not be written, is now marked incomplete: it is kept for you to look at, but it is not counted toward how many copies are kept and it is not sent to the off-site server. The run is reported as failed and says which account and what was missing. Before, such a run finished green, and a few bad nights in a row could push every good copy out.
- One account that fails no longer makes the whole nightly backup repeat every hour. Only that account is tried again (up to three times, then it is reported), and a run that is stopped part-way carries on where it stopped instead of starting again from the first account - so the accounts at the end of a long run are no longer left out. Only the two newest incomplete copies of an account are kept, so an account that fails every night cannot fill the disk.
- A weekly check now reads the backups back and verifies them: the structure of the local and off-site repositories, and a different twentieth of the local data each week, so everything has been read back within twenty weeks. It changes nothing and tells you if anything is wrong. It can also be run by hand from the command line. While it runs the backup repositories are busy, so backups wait for it (it is stopped after four hours at the latest, and only starts when a run has nothing left to back up). It is not started automatically on a server with an hourly backup schedule; run it by hand at a quiet time there. If it is stopped before it finishes you are told.
- Database copies are now compressed in a way that lets the incremental backup store only what changed, instead of a whole new copy of each database every night.
- When both a dated copy and a local incremental copy are made, each database is now dumped once, not twice.
- A backup that failed outright can no longer be mistaken for one that only skipped a few unreadable files.
- The backup log now says when a run had to wait for the repository, and how long.
- Optional, off by default: the nightly run can back up several accounts at once on a server with fast disks and spare processor time.
v0.1.251
4 October 2026
- Faster backups to a remote server on servers with many accounts: old copies are now cleaned up once at the end of a run instead of after every single account. Each clean-up locked the whole backup repository, so on a server with hundreds of accounts the clean-ups held up everything else. The disk space from old copies is freed at the end of the run.
v0.1.250
4 October 2026
- Fixed: the Backup page could take several minutes to open, or fail, while a backup clean-up was running on the off-site server. It no longer waits for the off-site server at all: it shows the list it already has, and refreshes it in the background. When nothing is remembered yet it shows the copies on this server straight away and says the off-site ones are still loading.
- A refresh now stops after a few minutes instead of running on, and only one runs at a time, so reloading the page can no longer pile up slow requests to the backup server.
v0.1.249
4 October 2026
- Smaller backups: the website files and mail in dated backups are now compressed with zstd instead of gzip. On a real account that was about a third smaller for website files and a fifth smaller for mail, for roughly the same time. Backups made before this still restore exactly as they did.
- A backup downloaded from the panel still comes as one .tar.gz, but the website and mail archives inside are now .zst files. They open with 7-Zip 23 or newer, or the zstd tool.
- Moving an account between two servers needs both on this version or later, because an older server cannot read a zstd backup.
- The File Manager can now unpack .tar.zst and single .zst files, the same way it already unpacks .tar.gz and .gz.
v0.1.248
4 October 2026
- Fixed: a Node.js application could occasionally fail to answer a request (the website showed a timeout) because the firewall rules that keep other accounts away from the application's port also dropped the harmless closing packets of connections already allowed. Those are now accepted; other accounts are still kept out.
Page 4 of 5
Want to see it running for yourself?
Get in touch and we'll walk you through it.