Skip to Content

Doctor and version

zekno doctor checks this machine and each server; zekno version says which release this is.

Pre-alpha and unaudited

The command-line tool, the runner and the core library have had internal reviews only. Do not use them for anything you could not afford to lose, and do not rely on them in production before an external audit.

zekno doctor

Checks this machine and each connected server, one line per check, and says what to do for each thing that is wrong. zekno itself changes nothing, on your machine or on any server. SSH keeps its own records: it may add the server to known_hosts on your machine, and the server logs the login.

Every check passingzekno doctor

zekno doctor with every check good. On this machine: zekno and the runner it installs, the key in the system keychain, the profile verified, and this folder linked to shop on prod-1. On prod-1: reached over SSH, Docker with Compose, the runner, and the free space on its disk. Then: no problems, no warnings.

zekno doctor with every check passing: this machine, then prod-1. The example after it has a warning and a problem.
$ zekno doctor
  This machine
✔ zekno          0.1.0, installs runner 0.1.0
✔ Key            in your system keychain
✔ Profile        verified, team list v2
✔ This folder    linked to demo on prod, compose.yaml is readable

  prod  deploy@203.0.113.7
✔ Reachable      over SSH
✔ Docker         Compose 5.6.0
! Runner         0.1.0-dev.1, older than this zekno's 0.1.0
                 Run zekno server update prod
✔ Disk           133 GB free

  staging  deploy@198.51.100.4
✖ Reachable      SSH needs an answer
                 Open zekno and press enter on staging
–  Docker        not checked
–  Runner        not checked
–  Disk          not checked

  1 problem, 1 warning.
  • This machine. zekno's version and the runner release it installs; the key (in the system keychain, and the one the profile names); the profile (it verifies, and its team list version); and this folder, when it has a zekno.json or a compose.yaml (the link is signed by your team and used here, and the Compose file can be read). It does not check the settings in a zekno.app.json: zekno settings does.

  • Each server. Whether it can be reached; Docker with Compose, and its version; the runner, and how its version stands to this zekno's; and the free disk space where the runner keeps its state. It finds these with a short fixed script that only reads, and by asking the runner its version and status.

  • The proxy. For a server whose runner speaks protocol 4 and holds a domain, one more line after the runner's: whether the proxy that answers for the domains runs (see Domains and HTTPS). A proxy that is missing or stopped is a problem. One from another image than the runner's, running or not, is a warning: the next deploy with a domain makes it again, and so does zekno server update when it replaces the runner. A server with no domain has no Proxy line.

    ✔ Proxy          answers for 2 domains
    ✖ Proxy          not running
                     Deploy an app with a domain to start it again
    ! Proxy          an older version
                     Deploy an app with a domain to replace it

    It does not check whether ports 80 and 443 are free: a deploy that starts the proxy says so when one is taken.

  • Results. ✔ good, ! a warning, ✖ a problem, – not checked: a check that depends on one that failed is not checked, not wrong. Every warning and problem has a line under it that says what to do. A warning does not fail the command; only a problem does, with exit status 1.

  • No questions. It reaches servers in SSH's batch mode and never asks for anything, so it never waits on a password. A server where SSH would have to ask says "SSH needs an answer": open the app with zekno and press enter on that server, where SSH can ask.

  • zekno changes nothing. zekno writes no file on your machine, links.json included; it writes, installs and starts nothing on a server; no deploy number is used. SSH's own records are as said above.

  • --server <name> checks your machine and that one server. --json prints one object: zekno, runner, machine (the checks of your machine), servers (each with name, where and checks), serversSkipped, latest, problems and warnings. Each check has check (the proxy's is proxy), result (good, warning, problem or not_checked), and detail and next where there are any.

  • It also asks GitHub once whether a newer release exists, as zekno version does. A newer one makes the zekno line a warning, with "A newer release exists" under it; no answer is not a warning.

zekno doctor --server prod

zekno version

Prints the version of this zekno and of the runner release it installs. zekno --version prints the same. With --json, one object: {"zekno":"0.1.0","runner":"0.1.0"}. It needs no key, no profile and no terminal, and changes nothing.

zekno version then asks GitHub once whether a newer release exists, and says so in one more line:

$ zekno version
zekno 0.1.0
runner 0.1.0, the runner release this zekno installs
✔ This is the newest release.

When a newer release exists, the last line is a warning instead: "! A newer release exists: 0.1.1. Update zekno to get it." When it could not find out, it is a dim Could not check for a newer release.: no network, no answer within 3 seconds, or an answer that is not a version. The exit status is 0 in each case. With --json the answer is the field latest, whose state is newest, newer (with the version) or not_checked: {"zekno":"0.1.0","runner":"0.1.0","latest":{"state":"newer","version":"0.1.1"}}.

  • It reads one small file from raw.githubusercontent.com, GitHub's host for files in a repository: the runner release file on the main branch of zekno-app/zekno, within 3 seconds. A newer version counts only once both its checksums are in that file, so a release that cannot be installed yet is not announced. zekno --version never uses the network.
  • The request carries the name zekno and nothing about you, your team, your servers or your project. GitHub learns the address and the time of the request.
  • Only zekno version and zekno doctor contact GitHub, at raw.githubusercontent.com, each once per run. Nothing else in zekno does so on its own, there is no check in the background, and nothing is stored.
  • zekno never downloads or installs a newer release. Until zekno is packaged, updating it means getting the newer source and building it again: see Build from source.