Repository navigation
HostZit 0.9.5
HostZit 0.9.5 — fixes from a full production test on real servers (h1.maxzit.com with two DNS-only servers): DNSSEC with DNS-only servers, buttons that did nothing in Chrome, SSH algorithms on servers installed by script, long malware scans and more.
Fixed
- DNSSEC with DNS-only or other secondary servers: a signed zone was never transferred to them again (NSEC3 "narrow" cannot be transferred), and with the DS record at the registrar validating resolvers rejected their answers. Zones are now signed with plain NSEC3 (RFC 9276) and reach the secondaries already signed; switching DNSSEC on or off tells them at once. Zones signed by older versions are changed by the update itself.
- "Open the customer panel" and "Open" of a restore point in the admin panel did nothing in Chrome, and so would the EmailZit sign-in through HostZit for a customer who still had to sign in (the browser blocked the redirect after the form).
- SSH and SFTP on servers installed by the install script offered weak algorithms (NIST curves, ECDSA host keys, SHA-1 and 64-bit MACs). Every server now uses the same strong set as the appliance.
- Malware scan of a large account was stopped after 30 minutes. A scan may now run up to 6 hours at the lowest priority.
- Log rotation: the package installed its files writable by the group, and logrotate refused HostZit's file, so website logs were never rotated. All files have proper permissions; the update fixes the installed file.
- DNS-only servers: change notices went to their IPv6 addresses too, and the transfer that followed was refused when only their IPv4 addresses were entered; only the entered servers are notified now. Running the DNS-only installer a second time (e.g. to add the primary's IPv6 address) stopped half-way; it now finishes and updates the zones it already follows.
- Mail from websites without a mail server stayed in the queue for ever without a word. The admin overview now says that websites cannot send mail, and waiting messages are dropped after 48 hours.
- REST API: a customer created with a domain below another customer's DNS zone got that website; the API answers 409
domain_takennow, like the customer panel. - WordPress staging: deleting a staging copy left its DNS record behind.
- Restore points in a folder of the server: every customer's restore point page showed the error "choose SFTP or S3".
- Certificates: the check whether a name points to the server read /etc/hosts first, where Debian (and cloud-init at every boot) maps the server's own name to 127.0.1.1, and could block the panel certificate.
- Address lists (website access, blocked addresses, DNS secondaries) accepted IPv6 addresses with a zone such as
fe80::1%eth0.
Updating
Settings → Updates installs 0.9.5. New servers:
curl -fsSL https://github.com/Max-Zit/hostzit/releases/latest/download/install.sh | sudo bash
DNS-only servers set up with an older installer: run the command shown under Settings → Hosting again (it lists both addresses of the HostZit server) and enter the DNS-only servers' IPv4 and IPv6 addresses under Secondary name servers.
Appliance images: hostzit-appliance.ova, .qcow2, -vhdx.zip, .iso (checksums in SHA256SUMS-appliance).