Dashboard – Hatoba

Dashboard at the top of the sidebar shows the online status and resource usage of every saved host on one page, without opening a tab on each. It reads Linux servers over SSH, as the terminal’s resource usage does (see Terminal): nothing is installed on the server and no Hatoba server is involved. The readings stay in memory on this device; they are never stored or synced.

English	简体中文	日本語
Dashboard	仪表盘	ダッシュボード
Cards	卡片	カード
Table	表格	表
Columns	列	列
Host List Order	主机列表顺序	ホスト一覧の順序
No Group	未分组	グループなし
Open Terminal	打开终端	ターミナルを開く
Filesystems	文件系统	ファイルシステム

The page

The header has a filter (All Hosts, Favorites, a group or a tag, as in the sidebar), the sort order, and Cards or Table. The page shows a section per group, in the sidebar’s order, then the hosts without a group.
A card shows the name, user@address:port, tags, the OS icon with a status badge, and CPU, memory and the root filesystem as bars, then the network rates, the load averages and the uptime. Every other filesystem over the disk threshold is named under the bars.
The table has a row per host. Columns turns each column (status, CPU, memory, disk, network, load, uptime) on or off. Pointing at the disk cell lists the filesystems.
Sort by the host list’s order (Host List Order), name, status (problems first), CPU, memory, disk (the fullest filesystem), network or uptime. Values sort highest first.
A value over its threshold shows in red. The thresholds are in Settings → General (90% by default; Off turns one off).
A host that stops answering keeps its last readings, dimmed, with the time they were read.
Click a card or row (or Tab to its name and press Enter) for the host’s detail: CPU and network over the last 3 minutes, memory and swap, the filesystems, the total traffic of the counted interfaces (since each came up, usually at boot), the load and uptime. Open Terminal brings the host’s tab to the front or opens one. Right-click a card or row for the same and Edit, which returns to the dashboard.
The filesystems are those that store files, at most 32: not memory-backed or kernel ones, container layers, read-only images such as snap packages, or the files a container runtime mounts, and each disk and each ZFS pool once. They are read every fifth reading.

How hosts are read

Hosts are read only while the page is shown, the window is visible and the vault is unlocked. Leaving the page stops reading, and the connections close a minute later, so a look at a tab and back does not sign in again. Locking closes them at once.
Each host is read every 3 seconds by default (Refresh interval in Settings → General), over a connected tab’s connection or else over a background SSH connection of its own. At most 8 connections are being set up at once, which keeps many hosts behind one jump host under OpenSSH’s default MaxStartups. Until a jump host has let one connection through, the hosts behind it try it one at a time.
A host whose first hop (the host, or its outermost jump host) does not answer the online probe is shown Offline and is not signed in to until it answers; it is checked again every minute, or at once with Retry in its menu.
When a host’s connection settings change (on another device, say), the dashboard connects again with the new ones and drops the readings made with the old ones.
Opening a tab on a host the dashboard is connected to takes over that connection: the tab opens its shell without signing in again, then sets the environment variables and starts the forwards set to start with the connection. Closing the tab closes the connection, and the dashboard connects again.
Servers that do not run Linux show their online status only. The dashboard signs in once to find that out, and again only after the vault was locked or the host’s connection settings changed.

English	简体中文	日本語
Online	在线	オンライン
Offline	离线	オフライン
Waiting	等待中	待機中
Read over an open terminal	通过已打开的终端读取	開いているターミナル経由で読み取り
Refresh interval	刷新间隔	更新間隔
Highlight CPU usage above	CPU 占用高亮阈值	CPU 使用率の強調しきい値
Highlight memory usage above	内存占用高亮阈值	メモリ使用率の強調しきい値
Highlight disk usage above	磁盘占用高亮阈值	ディスク使用率の強調しきい値
Off	关闭	オフ

Hosts that need the user

The background never shows a prompt. A host that cannot sign in on its own says why, with Connect, which opens a tab where the prompt can be answered. Once the tab is connected, the dashboard reads over it at once.

English	简体中文	日本語
Asks for a password on every connection	每次连接都询问密码	接続のたびにパスワードを確認します
The key’s passphrase isn’t saved	密钥口令未保存	鍵のパスフレーズが保存されていません
Asks for a password or verification code	需要输入密码或验证码	パスワードまたは確認コードの入力が必要です
Host key not confirmed yet	主机指纹尚未确认	ホスト鍵がまだ確認されていません
Host key changed	主机指纹已变更	ホスト鍵が変更されています
Sign-in failed	登录失败	サインインに失敗しました
Connect	连接	接続
Retry	重试	再試行

A failed sign-in, with the server’s answer, is not tried again on its own until the vault is locked and unlocked, so tools like fail2ban do not ban this computer. That includes a server that hangs up during sign-in, as one does after too many keys from the agent. Retry tries once more. Changing how the host connects (a new password, key, user, address or jump host, say) or confirming its host key in a tab also lets the dashboard try again; renaming it or marking it a favorite does not.
When a jump host turns one host away (or turns the dashboard away when the jump host itself is on it), every host behind it waits too, so its rejected password is not tried once per host. Retry on any of them tries again from that host only; the others follow once it gets past the jump host.
Other failures, such as a timeout or a lost connection, show the error and are tried again after a minute; Retry tries at once.

Leaving hosts out

Right-click a card for Leave Out of Dashboard, or use the … menu of a group’s section for Leave Group Out of Dashboard. A host left out is not shown or signed in to. The note at the bottom of the page counts the hosts left out; Show lists them dimmed, with Include in Dashboard, or Include Group in Dashboard for a host whose group is left out. The choice is kept on this device only and is not synced.

English	简体中文	日本語
Leave Out of Dashboard	不在仪表盘中显示	ダッシュボードから除外
Include in Dashboard	在仪表盘中显示	ダッシュボードに含める
Leave Group Out of Dashboard	不在仪表盘中显示此分组	グループをダッシュボードから除外
Include Group in Dashboard	在仪表盘中显示此分组	グループをダッシュボードに含める
Show	显示	表示
Left Out	未纳入	除外中

Not on the dashboard yet

History beyond the last few minutes, notifications, billing and traffic per billing cycle, temperatures, GPUs, Docker containers, custom commands per host, and a card for this computer. Readings stop when Hatoba is closed or the page is left: nothing is left running on the server.