REST API should support listing all units, including those that are stopped or failed about geard HOT 7 CLOSED

I'm pretty sure this is possible with /containers?all=1 request.

from geard.

@wallrj itamarst is correct, all=1 is the equivalent.

from geard.

Thanks @smarterclayton and sorry for the noise. Here's where I've started using the ?all=1 query string in our product.

• ClusterHQ/flocker#262 (comment)

I'd really appreciate if you had time to comment there about the different unit states that we can expect to be included by all=1. Also if you could comment on the circumstances in which a unit reaches the dead state. Or just some pointers to the relevant source code.

Thanks again for gear 😄

from geard.

By definition anything systemd is tracking in memory is in one of these states - with the special caveat that systemd reserves the right to remove anything from memory that it doesn't need to track. So http://cgit.freedesktop.org/systemd/systemd/tree/src/core/unit.h#n47 are the states systemd tracks. If something is unloaded it just means systemd doesn't need it. So anything that is linked into systemd, but not loaded in memory, has the equivalent ActiveState "inactive", LoadState "loaded", and SubState "dead", even though systemd doesn't return them.

from geard.

So geard provides the extra benefit of adding in stuff that is not in systemd memory by scanning the filesystem, as mentioned in the mailing list post above?

from geard.

Right - you can call systemd list-unit-files which scans all the watched folders - but that's a superset of what geard scans. Since geard only exposes certain named units and there is a convention for those, we settled for the slightly more efficient scan.

from geard.

Excellent, thanks.

from geard.