How it works
Dashboard and monitoring
What the landing page shows after sign-in, how it says a part could not be read, and the three disk figures that always add up to the size of the disk.
Every sign-in lands on the dashboard, System status. It answers three questions on one screen: what the server is doing, what is on it and what needs a decision.
What the dashboard shows
For an administrator the page has these parts:
- Right now: processor, memory and disk in use, and the load average over 1, 5 and 15 minutes, read live from the server agent.
- Services: whether the web server, PHP-FPM, the database server, the FTP server, cron and SSH are running.
- On this server: how many accounts, sites and databases it holds.
- Needs attention: certificates expiring soon, backups that failed, background tasks that failed, addresses the firewall has banned and a server clock that is not synchronised.
- Latest activity: the five newest entries of the audit journal, with a link to the whole journal.
- Quick actions: create an account, add a site, create a database, open the backups.
A customer sees their own account instead: links to their account, sites, databases and backups. None of the server-wide figures is sent to a customer.
A part that could not be read says so
The dashboard collects its parts from several modules, and any one of them can fail to answer. A part that could not be read is never shown as an empty list or a zero. "0 backups failed" and "Nothing needs attention" would be claims the panel could not back up.
- A count it could not take reads Unavailable.
- Needs attention names each source it could not check ("Backups could not be checked") and shows "Nothing needs attention" only when every source was read.
- The live reading, the services and the latest activity each say in their own card that they could not be read. They do not disappear.
For integrations, GET /api/v1/dashboard sends every section as {"state":"loaded","data":…} or
{"state":"unavailable"}. The response says whom it was built for in audience, either
administrator or customer; a customer's response carries no sections at all.
The disk: in use, reserved, available
The dashboard's disk tile and the disk card under Monitoring split the root filesystem into three parts in one bar. The three always add up to the size of the disk:
| Figure | What it is | df column |
|---|---|---|
| In use | the space the filesystem counts as occupied | Used |
| Reserved for the system | free space only root may write: 5% of an ext4 filesystem by default, 0 on xfs | none (Size - Used - Avail) |
| Available | what the hosting accounts can still write | Avail |
Earlier releases called "capacity minus available" the space in use. That counted the reserve as
used: on a 97.9 GiB ext4 disk the panel read 12.7 GiB where df said 7.7 GiB. In use is now
exactly df's figure.
The disk chart plots the same "in use" figure. Readings taken before the update stored the old figure, so they show as a gap until they leave the seven-day history.
The disk-full alert still counts the reserve as full. It fires when the accounts run out of room, because at that moment every account's write fails even though "in use" alone still reads about 95%.
For integrations, GET /api/v1/monitoring/metrics and the dashboard's resources carry
diskUsedBytes (df's Used), diskReservedBytes, diskAvailableBytes and diskTotalBytes.
Disk use by account
The table under Monitoring → Disk use by account sets each account's use against its plan's allowance. A row carries one of three badges when the allowance is not a live limit:
- Not enforced: the filesystem holding the homes cannot back the allowance with a disk limit, and the row says why (the filesystem has no quota accounting, or accounting is not switched on).
- Not checked: the server did not answer whether the allowance can be enforced, so the panel does not say either way. Try again shortly; if it persists, check that the agent is running.
- No badge: the allowance is enforced by the kernel.
For integrations the same three cases are the backing of each row of
GET /api/v1/monitoring/accounts-disk: enforced, notEnforced or unmeasured.
An earlier release showed "Not checked" rows as "Not enforced", which told an operator the allowance was not held, when in fact nobody had asked.
Alerts reach the operator by mail
Monitoring alerts and backup alerts are mailed to the address set under Outgoing mail → Send alerts to, through the same mail server. Both kinds go through one shared path. Each is written to the audit journal as "Alert raised", and with no alert address configured the journal also records "Mail skipped: no SMTP settings". What the backups mail about is on the file transfer and backups page.
Panel log level
Monitoring → Panel log level sets how much the panel writes to its log without a restart, from Verbose to Fatal. The level is the panel's own. The libraries it is built on (the message bus, the database driver, the HTTP client and the web platform) are always written at Warning or above, so choosing Verbose makes the panel log more but does not write every query or message into the journal, and choosing Error silences their warnings too. The platform's start-up and shutdown lines stay at Information. No level makes the panel log a password, a key or a token.