diff --git a/README.md b/README.md index 15bf5500fa200372af300126ae5a0e0cdbb1a537..0809ed1ccb1ca593a7b82d99a063658f990875f2 100644 --- a/README.md +++ b/README.md @@ -33,14 +33,14 @@ uncomment the `build` lines in the `docker-compose.yml`. The monitoring system can be started using - docker-compose up -d + docker compose up -d This will download and build all the required images and launch the containers for each service. It will also create an overlay network. To stop it it - docker-compose down + docker compose down ## Monitoring the monitoring @@ -49,7 +49,7 @@ Log files are kept in `logs/`, data dumps in `data/` and plots in `plots/`. To check the logs or follow them in real-time (`-f`) and limit the rewind to a number of lines `--tail=N`, e.g. - docker-compose logs -f --tail=10 SERVICE_NAME + docker compose logs -f --tail=10 SERVICE_NAME The `SERVICE_NAME` can be any of `backend`, `frontend`, `ligier`, `ligiermirror`, `ligierlogmirror`, `reco` or `livelog`. @@ -144,9 +144,9 @@ The monitoring processes talk to the KM3NeT Oracle DB service and need an valid session cookie. The monitoring servers of the ORCA and ARCA shore stations are whitelisted, but if you run the system on other machines, you need to provide the cookie string for that specific machine. To get the cookie string, run the -monitoring system with `docker-compose up -d` and connect to the backend with +monitoring system with `docker compose up -d` and connect to the backend with - # docker exec -it monitoring_backend_1 bash + # docker exec -it monitoring-backend-1 bash To get a session cookie, query the database however you like, e.g. @@ -165,5 +165,5 @@ container!) with the following entry (of course with your cookie string): and restart the whole monitoring system with - docker-compose down && docker-compose up -d + docker compose down && docker compose up -d diff --git a/backend/supervisord.conf.example b/backend/supervisord.conf.example index e2cc3cdc69ef1a648dbaa96a131439135de533dc..03cde814bbda3fbd9feeb946f1a7eebaa0632ec0 100644 --- a/backend/supervisord.conf.example +++ b/backend/supervisord.conf.example @@ -37,7 +37,7 @@ stdout_logfile=/logs/%(program_name)s.out.log ; stdout log path, NONE for stderr_logfile=/logs/%(program_name)s.err.log ; stderr log path, NONE for none; default AUTO [program:ahrs_calibration] -command=python -u scripts/ahrs_calibration.py -d %(ENV_DETECTOR_ID)s -l monitoring_ligier_1 +command=python -u scripts/ahrs_calibration.py -d %(ENV_DETECTOR_ID)s -l monitoring-ligier-1 ;process_name=%(program_name)s ; process_name expr (default %(program_name)s) ;priority=999 ; the relative start priority (default 999) startsecs=5 ; # of secs prog must stay up to be running (def. 1) @@ -48,7 +48,7 @@ stdout_logfile=/logs/%(program_name)s.out.log ; stdout log path, NONE for stderr_logfile=/logs/%(program_name)s.err.log ; stderr log path, NONE for none; default AUTO [program:dom_activity] -command=python -u scripts/dom_activity.py -d %(ENV_DETECTOR_ID)s -l monitoring_ligier_1 +command=python -u scripts/dom_activity.py -d %(ENV_DETECTOR_ID)s -l monitoring-ligier-1 startsecs=5 ; # of secs prog must stay up to be running (def. 1) startretries=30 ; max # of serial start failures when starting (default 3) autorestart=true ; when to restart if exited after running (def: unexpected) @@ -56,7 +56,7 @@ stdout_logfile=/logs/%(program_name)s.out.log stderr_logfile=/logs/%(program_name)s.err.log [program:dom_rates] -command=python -u scripts/dom_rates.py -d %(ENV_DETECTOR_ID)s -l monitoring_ligier_1 +command=python -u scripts/dom_rates.py -d %(ENV_DETECTOR_ID)s -l monitoring-ligier-1 startsecs=5 ; # of secs prog must stay up to be running (def. 1) startretries=30 ; max # of serial start failures when starting (default 3) autorestart=true ; when to restart if exited after running (def: unexpected) @@ -64,7 +64,7 @@ stdout_logfile=/logs/%(program_name)s.out.log stderr_logfile=/logs/%(program_name)s.err.log [program:pmt_rates] -command=python -u scripts/pmt_rates.py -d %(ENV_DETECTOR_ID)s -l monitoring_ligier_1 -i 20 -u 1 +command=python -u scripts/pmt_rates.py -d %(ENV_DETECTOR_ID)s -l monitoring-ligier-1 -i 20 -u 1 startsecs=5 ; # of secs prog must stay up to be running (def. 1) startretries=30 ; max # of serial start failures when starting (default 3) autorestart=true ; when to restart if exited after running (def: unexpected) @@ -72,7 +72,7 @@ stdout_logfile=/logs/%(program_name)s.out.log stderr_logfile=/logs/%(program_name)s.err.log [program:trigger_rates] -command=python -u scripts/trigger_rates.py -l monitoring_ligier_1 +command=python -u scripts/trigger_rates.py -l monitoring-ligier-1 startsecs=5 ; # of secs prog must stay up to be running (def. 1) startretries=30 ; max # of serial start failures when starting (default 3) autorestart=true ; when to restart if exited after running (def: unexpected) @@ -80,7 +80,7 @@ stdout_logfile=/logs/%(program_name)s.out.log stderr_logfile=/logs/%(program_name)s.err.log [program:triggermap] -command=python -u scripts/live_triggermap.py -d %(ENV_DETECTOR_ID)s -l monitoring_ligier_1 +command=python -u scripts/live_triggermap.py -d %(ENV_DETECTOR_ID)s -l monitoring-ligier-1 startsecs=5 ; # of secs prog must stay up to be running (def. 1) startretries=30 ; max # of serial start failures when starting (default 3) autorestart=true ; when to restart if exited after running (def: unexpected) @@ -88,7 +88,7 @@ stdout_logfile=/logs/%(program_name)s.out.log stderr_logfile=/logs/%(program_name)s.err.log [program:ztplot] -command=python -u scripts/ztplot.py -d %(ENV_DETECTOR_ID)s -l monitoring_ligier_1 +command=python -u scripts/ztplot.py -d %(ENV_DETECTOR_ID)s -l monitoring-ligier-1 startsecs=5 ; # of secs prog must stay up to be running (def. 1) startretries=30 ; max # of serial start failures when starting (default 3) autorestart=true ; when to restart if exited after running (def: unexpected) @@ -104,7 +104,7 @@ stdout_logfile=/logs/%(program_name)s.out.log stderr_logfile=/logs/%(program_name)s.err.log [program:triggeredhits] -command=python -u scripts/triggered_hits.py -d %(ENV_DETECTOR_ID)s -l monitoring_ligier_1 +command=python -u scripts/triggered_hits.py -d %(ENV_DETECTOR_ID)s -l monitoring-ligier-1 startsecs=5 ; # of secs prog must stay up to be running (def. 1) startretries=30 ; max # of serial start failures when starting (default 3) autorestart=true ; when to restart if exited after running (def: unexpected) @@ -141,7 +141,7 @@ stderr_logfile=/logs/%(program_name)s.err.log ;stderr_logfile=/logs/%(program_name)s.err.log [program:timesync_monitor] -command=python -u scripts/timesync_monitor.py -l monitoring_ligier_1 -m %(ENV_LOG_LIGIER_IP)s -q %(ENV_LOG_LIGIER_PORT)s +command=python -u scripts/timesync_monitor.py -l monitoring-ligier-1 -m %(ENV_LOG_LIGIER_IP)s -q %(ENV_LOG_LIGIER_PORT)s startsecs=5 ; # of secs prog must stay up to be running (def. 1) startretries=30 ; max # of serial start failures when starting (default 3) autorestart=true ; when to restart if exited after running (def: unexpected) diff --git a/docker-compose.yml b/docker-compose.yml index 4357eb6f0d516fec0037823c5e1a2e58d227c6b0..8ad0bbdaf3c37925d880333035745cfa5048d259 100644 --- a/docker-compose.yml +++ b/docker-compose.yml @@ -1,7 +1,10 @@ -version: "3" services: ligier: image: "docker.km3net.de/jpp:v14.4.3" + deploy: + resources: + limits: + memory: 1024M command: JLigier -d2 volumes: - ./logs:/logs @@ -10,9 +13,13 @@ services: ligiermirror: image: "docker.km3net.de/jpp:v14.4.3" + deploy: + resources: + limits: + memory: 512M env_file: - .env - command: JLigierMirror -H ${DAQ_LIGIER_IP}:${DAQ_LIGIER_PORT} -X monitoring_ligier_1:5553 -t "${TAGS_TO_MIRROR}" + command: JLigierMirror -H ${DAQ_LIGIER_IP}:${DAQ_LIGIER_PORT} -X monitoring-ligier-1:5553 -t "${TAGS_TO_MIRROR}" volumes: - ./logs:/logs depends_on: @@ -21,9 +28,13 @@ services: ligierlogmirror: image: "docker.km3net.de/jpp:v14.4.3" + deploy: + resources: + limits: + memory: 512M env_file: - .env - command: JLigierMirror -H ${LOG_LIGIER_IP}:${LOG_LIGIER_PORT} -X monitoring_ligier_1:5553 -t "MSG Born Died" + command: JLigierMirror -H ${LOG_LIGIER_IP}:${LOG_LIGIER_PORT} -X monitoring-ligier-1:5553 -t "MSG Born Died" volumes: - ./logs:/logs depends_on: @@ -32,9 +43,13 @@ services: ligiertmchmirror: image: "docker.km3net.de/jpp:v14.4.3" + deploy: + resources: + limits: + memory: 512M env_file: - .env - command: JLigierMirror -H ${TMCH_LIGIER_IP}:${TMCH_LIGIER_PORT} -X monitoring_ligier_1:5553 -t IO_MONIT + command: JLigierMirror -H ${TMCH_LIGIER_IP}:${TMCH_LIGIER_PORT} -X monitoring-ligier-1:5553 -t IO_MONIT volumes: - ./logs:/logs depends_on: @@ -44,6 +59,10 @@ services: backend: # build: ./backend image: "docker.km3net.de/km3mon-backend:${KM3MON_VERSION}" + deploy: + resources: + limits: + memory: 8192M env_file: - .env volumes: @@ -57,9 +76,13 @@ services: reco: # build: ./reco image: "docker.km3net.de/km3mon-reco:${KM3MON_VERSION}" + deploy: + resources: + limits: + memory: 4096M env_file: - .env - command: monitoring_ligier_1 ${MONITORING_LIGIER_PORT} + command: monitoring-ligier-1 ${MONITORING_LIGIER_PORT} volumes: - ./reco:/reco - ./logs:/logs @@ -71,6 +94,10 @@ services: frontend: # build: ./frontend image: "docker.km3net.de/km3mon-frontend:${KM3MON_VERSION}" + deploy: + resources: + limits: + memory: 1024M env_file: - .env ports: @@ -84,8 +111,28 @@ services: livelog: # build: ./livelog image: "docker.km3net.de/km3mon-livelog:${KM3MON_VERSION}" + deploy: + resources: + limits: + memory: 1024M ports: - ${LOGGING_PORT}:9001 volumes: - ./logs:/logs restart: unless-stopped + + fibonaccifit: + image: "fibonaccifit" + deploy: + resources: + limits: + memory: 4096M + env_file: + - .env + command: monitoring-ligier-1 ${MONITORING_LIGIER_PORT} + volumes: + - ./logs:/logs + - ./plots:/plots + - ./data:/data + - ./fibonaccifit:/reco/src + restart: unless-stopped