orcaLog in

Pipelines

mojo_embed

Progress

Runs at onceRuns of this pipeline going now, out of the most the scheduler starts at once (max_runs); the others wait in the queue. Fill runs come on top, up to fill_max_runs. Docs ↗
0 running of 1, 0 queued
Next runWhen the schedule fires next, in UTC, and the partition that run fills. Next runs, further down, lists the next five. Docs ↗
2026-11-11T07:00:00Z, fills 2026-11-11
CoverageHow much of the pipeline is done: its days or months done of those in its range so far (its first partition through the last that has ended), or a keyed pipeline's keys done of the keys its dataset has. Done means every step succeeded in a run of its current code, as on the calendar; failed, skipped, queued and running ones aren't. A dash means it has none to count yet. Docs ↗
1 / 1 day done (100%)

Config

PartitionsHow the work is split: each run does some partitions (months, days, hours or named keys), from the first one on. Backfills and the coverage calendar count in partitions. Docs ↗
daily from 2026-11-10
ScheduleThe cron line the scheduler fires on, in its time zone, and the partition each fire runs. None means runs start only from a sensor, a backfill or a person. Docs ↗
0 7 * * * UTC, fills the current partition
Run retriesHow many more times the scheduler starts a run that failed before leaving it failed for a person. A step's own retries, inside a run, come first. Docs ↗
3
PlacementWhere its runs go: control is the scheduler's own host, ec2 an instance launched for the run, worker:<pool> a remote orca worker of that pool, and default the scheduler's runner (ORCA_RUNNER). Docs ↗
default (the scheduler's runner)
BinaryThe program each run starts: the pipeline's Mojo code built into one executable, as orca register recorded it. Docs ↗
/opt/orca/pipelines/mojo_embed

CodeThe functions this pipeline is made of, at the commit it's registered with. Hover a name to see its source; click it to open the file on its host. Docs ↗

PipelineWhere the pipeline is declared: its Pipeline[...] line. Docs ↗
mojo_embed
StepsEach step's function, in the order the steps run. The Steps table below has what each reads and writes. Docs ↗
embed

Code hash 0xcb48f6988c24e2f3A hash of what the binary says about itself: its steps, partitions and schedule. It changes when they do, not on every build. Docs ↗

ResourcesWhat its latest runs used, as the median (p50) and the most of them: what a run of it needs. A run's peak is its attempts' largest, its CPU their sum. Docs ↗

MeasuredHow many of its latest runs these are: the last 20 that recorded what they used.
its last 1 run
Peak memoryThe most memory its processes held at once: the run's process and what it started (dista's local workers too), sampled every second by the scheduler or worker that started it, and at the end the largest single process. Not measured on EC2 or Kubernetes. Docs ↗
p50 441 MB · max 441 MB
CPU usedThe CPU time its processes used, user and system, and the most cores they kept busy over one sample's second. macOS gives only the time, and only for a process that made one attempt. Docs ↗
p50 63.2 s · max 63.2 s
Peak coresThe most cores a run's processes kept busy over one sample's second: the CPUs to ask for. Not measured on macOS. Docs ↗
p50 2.8 · max 2.8
RequestedWhat it asks for (cpus and memory_gb in orca.toml): its runs start once the least of it fits on their host, each granted up to the most. Set it from the peaks above. Docs ↗
memory auto..2 GB
LearnedThe request its latest successful runs' peaks suggest: the 95th percentile of their peak memory plus 15%, at least what an out-of-memory raised it to, and the 90th percentile of their peak cores, within its request's range. Applied where cpus or memory_gb is auto, unless learn = false. Docs ↗
not learned yet: 1 of the 3 successful measured runs it learns from

Steps docs

GraphTable
embed (view source)embedmapped ↗

Next runs docs

UTCWhen the schedule fires, in UTC.Wall clockThe same moment in the pipeline's time zone.FillsThe partition that run fills.
2026-11-11T07:00:00Z2026-11-11T07:00:002026-11-11
2026-11-12T07:00:00Z2026-11-12T07:00:002026-11-12
2026-11-13T07:00:00Z2026-11-13T07:00:002026-11-13
2026-11-14T07:00:00Z2026-11-14T07:00:002026-11-14
2026-11-15T07:00:00Z2026-11-15T07:00:002026-11-15

Configuration

Registered from /opt/orca/orca.toml @ 48c60575736b of https://github.com/magmalake/mojo-census, sha256 42bcd33a9baf.

KeyThe setting's name, as orca.toml and orca register's flags call it. Hover the name for what it sets.ValueThe value it is registered with.FromWhere the value came from: the code's default, orca.toml (and which table), a flag of orca register, or the scheduler's environment. A link opens the line that sets it.
placementWhere its runs start: default (the scheduler's runner), control, ec2, or worker:<pool>.defaultcode default the scheduler's runner, local
head_typeThe EC2 instance type of a run's dista head.c7g.largecode default the ec2 runner's
worker_typeThe EC2 instance type of each dista worker.c7g.xlargecode default the ec2 runner's
workersHow many dista workers a run asks for, or min..max to scale between.0code default the pipeline has no tasks
worker_procsDista worker processes on each worker instance; auto for one per core.1code default
max_runsThe most runs of the pipeline the scheduler starts at once.1orca.toml [defaults]
memory_gbauto..2orca.toml [pipelines.mojo_embed]
fillsall: the scheduler fills the whole range by itself. since_registration: from registration on.since_registrationcode default
env.MOJO_EMBED_MODELA variable set in its runs' environment./tmp/mojo-census/models/bge-small-en-v1.5orca.toml [pipelines.mojo_embed]
env.ORCA_PUBLIC_WAREHOUSEA variable set in its runs' environment.s3://census-mlake-dev/warehouseenvironment ORCA_RUN_ENV