ops: 0 compose service(s) drifted out of existence (#2747 watch) + 1 container(s) unhealthy for 1.0h+ (#3030 watch)

[github-actions]

Sweep at 2026-09-18T00:36:32Z on supermicro: 0 compose-defined, always-on service(s) have no container at all (not even stopped), 0 live stack(s) no longer exist in the repo manifest, 1 container(s) have had a healthcheck failing streak continuously for at least 1.0h, and 0 running container(s) have a cpus/memory limit that diverges from the repo's declared value.

Healthcheck failing streak

Context: #3030 — a container that exists and is Up but keeps failing its own healthcheck (FailingStreak from docker inspect) continuously for at least 1.0h. Tracked as a duration across sweeps, not a raw streak count — hp-autoheal restarts a container the moment Docker marks it unhealthy, which resets FailingStreak to 0 well before any fixed count threshold is reached (#3023 observed exactly FailingStreak=3, never higher). This is a distinct failure mode from the sections above: the container was never missing, so restart: policies never get a chance to help — the thing that's broken outlives a container restart (a wedged dependency, a resource the healthcheck needs that a plain restart doesn't free). Check docker inspect <name> --format '{{json .State.Health}}' for the actual probe failure before restarting anything.

• hp-arkime-capture — unhealthy for 2.3h (10 consecutive healthcheck failures, status: starting)

[github-actions]

Sweep at 2026-09-18T05:32:23Z on supermicro-ci-3: 0 compose-defined, always-on service(s) have no container at all (not even stopped), 0 live stack(s) no longer exist in the repo manifest, 1 container(s) have had a healthcheck failing streak continuously for at least 1.0h, and 0 running container(s) have a cpus/memory limit that diverges from the repo's declared value.

Healthcheck failing streak

Context: #3030 — a container that exists and is Up but keeps failing its own healthcheck (FailingStreak from docker inspect) continuously for at least 1.0h. Tracked as a duration across sweeps, not a raw streak count — hp-autoheal restarts a container the moment Docker marks it unhealthy, which resets FailingStreak to 0 well before any fixed count threshold is reached (#3023 observed exactly FailingStreak=3, never higher). This is a distinct failure mode from the sections above: the container was never missing, so restart: policies never get a chance to help — the thing that's broken outlives a container restart (a wedged dependency, a resource the healthcheck needs that a plain restart doesn't free). Check docker inspect <name> --format '{{json .State.Health}}' for the actual probe failure before restarting anything.

• hp-arkime-capture — unhealthy for 7.2h (3 consecutive healthcheck failures, status: starting)

[github-actions]

Sweep at 2026-09-18T10:32:27Z on supermicro-ci-3: 0 compose-defined, always-on service(s) have no container at all (not even stopped), 0 live stack(s) no longer exist in the repo manifest, 1 container(s) have had a healthcheck failing streak continuously for at least 1.0h, and 0 running container(s) have a cpus/memory limit that diverges from the repo's declared value.

Healthcheck failing streak

Context: #3030 — a container that exists and is Up but keeps failing its own healthcheck (FailingStreak from docker inspect) continuously for at least 1.0h. Tracked as a duration across sweeps, not a raw streak count — hp-autoheal restarts a container the moment Docker marks it unhealthy, which resets FailingStreak to 0 well before any fixed count threshold is reached (#3023 observed exactly FailingStreak=3, never higher). This is a distinct failure mode from the sections above: the container was never missing, so restart: policies never get a chance to help — the thing that's broken outlives a container restart (a wedged dependency, a resource the healthcheck needs that a plain restart doesn't free). Check docker inspect <name> --format '{{json .State.Health}}' for the actual probe failure before restarting anything.

• hp-arkime-capture — unhealthy for 12.2h (11 consecutive healthcheck failures, status: starting)

[github-actions]

Sweep at 2026-09-18T14:46:42Z on supermicro: 0 compose-defined, always-on service(s) have no container at all (not even stopped), 0 live stack(s) no longer exist in the repo manifest, 1 container(s) have had a healthcheck failing streak continuously for at least 1.0h, and 0 running container(s) have a cpus/memory limit that diverges from the repo's declared value.

Healthcheck failing streak

Context: #3030 — a container that exists and is Up but keeps failing its own healthcheck (FailingStreak from docker inspect) continuously for at least 1.0h. Tracked as a duration across sweeps, not a raw streak count — hp-autoheal restarts a container the moment Docker marks it unhealthy, which resets FailingStreak to 0 well before any fixed count threshold is reached (#3023 observed exactly FailingStreak=3, never higher). This is a distinct failure mode from the sections above: the container was never missing, so restart: policies never get a chance to help — the thing that's broken outlives a container restart (a wedged dependency, a resource the healthcheck needs that a plain restart doesn't free). Check docker inspect <name> --format '{{json .State.Health}}' for the actual probe failure before restarting anything.

• hp-arkime-capture — unhealthy for 16.4h (10 consecutive healthcheck failures, status: starting)

[github-actions]

Sweep at 2026-09-18T18:03:44Z on supermicro-ci-2: 0 compose-defined, always-on service(s) have no container at all (not even stopped), 0 live stack(s) no longer exist in the repo manifest, 1 container(s) have had a healthcheck failing streak continuously for at least 1.0h, and 0 running container(s) have a cpus/memory limit that diverges from the repo's declared value.

Healthcheck failing streak

Context: #3030 — a container that exists and is Up but keeps failing its own healthcheck (FailingStreak from docker inspect) continuously for at least 1.0h. Tracked as a duration across sweeps, not a raw streak count — hp-autoheal restarts a container the moment Docker marks it unhealthy, which resets FailingStreak to 0 well before any fixed count threshold is reached (#3023 observed exactly FailingStreak=3, never higher). This is a distinct failure mode from the sections above: the container was never missing, so restart: policies never get a chance to help — the thing that's broken outlives a container restart (a wedged dependency, a resource the healthcheck needs that a plain restart doesn't free). Check docker inspect <name> --format '{{json .State.Health}}' for the actual probe failure before restarting anything.

• hp-arkime-capture — unhealthy for 19.7h (10 consecutive healthcheck failures, status: starting)

[github-actions]

Sweep at 2026-09-18T21:06:28Z on supermicro-ci-4: 0 compose-defined, always-on service(s) have no container at all (not even stopped), 0 live stack(s) no longer exist in the repo manifest, 1 container(s) have had a healthcheck failing streak continuously for at least 1.0h, and 0 running container(s) have a cpus/memory limit that diverges from the repo's declared value.

Healthcheck failing streak

Context: #3030 — a container that exists and is Up but keeps failing its own healthcheck (FailingStreak from docker inspect) continuously for at least 1.0h. Tracked as a duration across sweeps, not a raw streak count — hp-autoheal restarts a container the moment Docker marks it unhealthy, which resets FailingStreak to 0 well before any fixed count threshold is reached (#3023 observed exactly FailingStreak=3, never higher). This is a distinct failure mode from the sections above: the container was never missing, so restart: policies never get a chance to help — the thing that's broken outlives a container restart (a wedged dependency, a resource the healthcheck needs that a plain restart doesn't free). Check docker inspect <name> --format '{{json .State.Health}}' for the actual probe failure before restarting anything.

• hp-arkime-capture — unhealthy for 22.8h (7 consecutive healthcheck failures, status: starting)

[github-actions]

Sweep at 2026-09-18T23:28:04Z on supermicro-ci-2: 0 compose-defined, always-on service(s) have no container at all (not even stopped), 0 live stack(s) no longer exist in the repo manifest, 1 container(s) have had a healthcheck failing streak continuously for at least 1.0h, and 0 running container(s) have a cpus/memory limit that diverges from the repo's declared value.

Healthcheck failing streak

Context: #3030 — a container that exists and is Up but keeps failing its own healthcheck (FailingStreak from docker inspect) continuously for at least 1.0h. Tracked as a duration across sweeps, not a raw streak count — hp-autoheal restarts a container the moment Docker marks it unhealthy, which resets FailingStreak to 0 well before any fixed count threshold is reached (#3023 observed exactly FailingStreak=3, never higher). This is a distinct failure mode from the sections above: the container was never missing, so restart: policies never get a chance to help — the thing that's broken outlives a container restart (a wedged dependency, a resource the healthcheck needs that a plain restart doesn't free). Check docker inspect <name> --format '{{json .State.Health}}' for the actual probe failure before restarting anything.

• hp-arkime-capture — unhealthy for 25.1h (10 consecutive healthcheck failures, status: starting)

[github-actions]

Sweep at 2026-09-19T01:27:52Z on supermicro-ci-4: 0 compose-defined, always-on service(s) have no container at all (not even stopped), 0 live stack(s) no longer exist in the repo manifest, 1 container(s) have had a healthcheck failing streak continuously for at least 1.0h, and 0 running container(s) have a cpus/memory limit that diverges from the repo's declared value.

Healthcheck failing streak

Context: #3030 — a container that exists and is Up but keeps failing its own healthcheck (FailingStreak from docker inspect) continuously for at least 1.0h. Tracked as a duration across sweeps, not a raw streak count — hp-autoheal restarts a container the moment Docker marks it unhealthy, which resets FailingStreak to 0 well before any fixed count threshold is reached (#3023 observed exactly FailingStreak=3, never higher). This is a distinct failure mode from the sections above: the container was never missing, so restart: policies never get a chance to help — the thing that's broken outlives a container restart (a wedged dependency, a resource the healthcheck needs that a plain restart doesn't free). Check docker inspect <name> --format '{{json .State.Health}}' for the actual probe failure before restarting anything.

• hp-arkime-capture — unhealthy for 27.1h (8 consecutive healthcheck failures, status: starting)

[github-actions]

Recovered as of 2026-09-19T06:23:56Z: sweep on supermicro found no drifted service. Closing; the next sweep that finds drift reopens.