I'm still under development. Some of what you see below is outdated, while other parts may be aspirational. Talk to Tom before doing anything that matters.
A Docker container to collect VIIRS data at AVO.
I'll do all of my work in /viirs. Mount something sensible there, I'll use about 36Gig per day. Data files, logs, and TLE files will be placed under here.
Daily I'll trim /viirs, removing all files more than $VIIRS_RETENTION days old. This is a working directory, put anything you want to keep somewhere else.
I will publish a message on port 29092/tcp for each file successfully downloaded. See msg_publisher for more details and an example client.
I will also distribute product creation tasks on port 19091/tcp and publish messages describing the queue of product creation tasks waiting to be produced on port 19191/tcp. See task_broker for more details.
Launches all other deamons and keeps them running. supervisord creates a log file for each daemon it spawns and places it in /viirs/log Unfortunatly, supervisord creates the logs with very restrictive perms. There's an open issue on it.
supervisor also writes its own log. Look here to find unstable daemons. This log should be short and boring.
Additional info on supervisord is available at http://supervisord.org/. the supervisord log to see how stable the daemons are. When everything is going well this is a short boring log.
Launch mirror_gina and cleanup. Additional info on supercronic is available at https://github.com/aptible/supercronic. The supercronic logs will capture anything interesting from mirror_gina.
nameserver is part of the PyTroll posttroll project. Additional info on posttrol is available at https://github.com/pytroll/posttroll. nameserver keeps track of topics learned from messages broadcasted by publishers and responds to queries from listeners looking for a topic. It has no configuration file and rarely causes trouble. This is how task_broker learns about running segment_gatherers.
kicks things off once a file has been downloaded. trollstalker is part of the PyTroll pytroll-collectors project. Additional info on pytroll- ollectors is available at https://github.com/pytroll/pytroll-collectors.
trollstalker uses inotify to watch for new files and publishes messages to start processing. This means things will only run on Linux. Additionally, it's important to think about inotify's scope. A NFS filesystem with a remote file writer won't work, as inotify wouldn't see the I/O.
listens to trollstalker and emits a message once enough data is available to produce a product. segment_gatherer is part of the PyTroll pytroll-collectors project. Additional info on pytroll-collectors is available at https://github.com/pytroll/pytroll-collectors.
One instance of segment_gatherer runs for each product.
collects messages from segment_gatherer and provides tasks to viirsprocessors.
Distribute product generation tasks. Tasks are encoded as UTF-8 encoded strings, suitable for parsing by posttroll.message.Message.decode(). task_broker tries to avoid assigning the same task twice. If a message arrives which matches one already in the queue, that queued task will maintain its queue postition and be updated with the new message. Duplicate tasks are determined with a 3-tupple of message.subject, message.platform_name, and message.orbit_number. tasks are further divided to seperate ascending and descending passes.
Publish messages generated by internal processing on port 29092/tcp. Most usefully this includes a message for each file downloaded. Messages are encoded as UTF-8 encoded strings, which can be unmarshalled with posttroll.message.Message.decode or parsed by hand. These messages are simple to parse, but the format is undocumented. I intend to revisit this format and formally document it in the future.
Example subscriber:
#!/usr/bin/env python3
import json
import zmq
from posttroll.message import Message
context = zmq.Context()
socket = context.socket(zmq.SUB)
socket.setsockopt_string(zmq.SUBSCRIBE, 'pytroll://AVO/viirs/sdr')
socket.connect("tcp://avoworker2:29092")
while True:
msg = Message.decode(socket.recv())
print(json.dumps(msg.data, sort_keys=True, indent=4, default=str))
Produces:
gilbert [6:59pm] % ./pubtest.py
{
"end_time": "2019-04-22 18:29:59",
"orbit_number": 38780,
"orig_platform_name": "npp",
"platform_name": "Suomi-NPP",
"proctime": "2019-04-22 18:39:26.285541",
"segment": "SVM16",
"sensor": [
"viirs"
],
"start_time": "2019-04-22 18:28:34.800000",
"uid": "SVM16_npp_d20190422_t1828348_e1829590_b38780_c20190422183926285541_cspp_dev.h5",
"uri": "/viirs/sdr/SVM16_npp_d20190422_t1828348_e1829590_b38780_c20190422183926285541_cspp_dev.h5"
}
Searches GINA NRT with parameters provided in the environemnt and retrieves any new files found. mirror_gina doesn't write its own logs, look for output in the supercronic logs.
Nightly I'll cleanup files in /viirs, removing any that are older than $VIIRS_RETENTION days. All files, even ones you might not think of.
- VIIRS_RETENTION - if this is defined, files in /viirs will be cleaned up after this many days.
- NUM_GINA_CONNECTIONS - files will be retrieved in segments using this many connections in parallel
- GINA_BACKFILL_DAYS - data will be made complete over this many days
- VIIRS_FACILITY - use data received at UAF or Gilmore Creek?
I will email errors generated by mirror_gina and task_broker if these are provided.
- MAILHOST - who can forward mail for me?
- LOG_SENDER - From: address
- LOG_RECIPIENT - To: address
Here is an example service stanza for use with docker-compose.
viirscollector:
image: "tparkerusgs/avoviirscollector:release-2.0.2"
ports:
- 19191:19191
- 19091:19091
- 29092:29092
user: "2001"
environment:
- VIIRS_RETENTION=7
- NUM_GINA_CONNECTIONS=4
- GINA_BACKFILL_DAYS=2
- VIIRS_FACILITY=gilmore
- MAILHOST=smtp.address.com
- LOG_SENDER=worker@address.com
- LOG_RECIPIENT=email@address.com
restart: always
logging:
driver: json-file
options:
max-size: 10m
volumes:
- type: volume
source: viirs
target: /viirs
volume:
nocopy: true