From 0df118f10575cfadfdd5fdd27855a27ce2007060 Mon Sep 17 00:00:00 2001 From: Tamas Gal <himself@tamasgal.com> Date: Wed, 28 Dec 2022 22:07:53 +0100 Subject: [PATCH] Add explicit restart policies in supervisord.conf --- backend/supervisord.conf.example | 51 ++++++++++++++++++++++++++++++-- 1 file changed, 48 insertions(+), 3 deletions(-) diff --git a/backend/supervisord.conf.example b/backend/supervisord.conf.example index 2a81a73..e2cc3cd 100644 --- a/backend/supervisord.conf.example +++ b/backend/supervisord.conf.example @@ -40,65 +40,98 @@ stderr_logfile=/logs/%(program_name)s.err.log ; stderr log path, NONE for 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=1 ; # of secs prog must stay up to be running (def. 1) -;startretries=3 ; max # of serial start failures when starting (default 3) -;autorestart=unexpected ; when to restart if exited after running (def: unexpected) +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) ;redirect_stderr=true ; redirect proc stderr to stdout (default false) stdout_logfile=/logs/%(program_name)s.out.log ; stdout log path, NONE for none; default AUTO 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 +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) 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 +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) 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 +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) 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 +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) 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 +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) 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 +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) stdout_logfile=/logs/%(program_name)s.out.log stderr_logfile=/logs/%(program_name)s.err.log [program:time_residuals] command=python -u scripts/time_residuals.py %(ENV_ROYFIT_TIMERES)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) 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 +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) stdout_logfile=/logs/%(program_name)s.out.log stderr_logfile=/logs/%(program_name)s.err.log [program:time_residuals] command=python -u scripts/time_residuals.py /data/reco_timeres.csv +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) stdout_logfile=/logs/%(program_name)s.out.log stderr_logfile=/logs/%(program_name)s.err.log [program:rttc] command=python -u scripts/rttc.py -d %(ENV_DETECTOR_ID)s -l %(ENV_DETECTOR_MANAGER_IP)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) stdout_logfile=/logs/%(program_name)s.out.log stderr_logfile=/logs/%(program_name)s.err.log [program:acoustics] command=python -u scripts/acoustics.py -d %(ENV_DETECTOR_ID)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) stdout_logfile=/logs/%(program_name)s.out.log stderr_logfile=/logs/%(program_name)s.err.log @@ -109,23 +142,35 @@ 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 +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) stdout_logfile=/logs/%(program_name)s.out.log stderr_logfile=/logs/%(program_name)s.err.log [program:msg_dumper] command=python -u scripts/msg_dumper.py -l %(ENV_LOG_LIGIER_IP)s -p %(ENV_LOG_LIGIER_PORT)s -o /logs +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) priority=200 stdout_logfile=/logs/%(program_name)s.out.log stderr_logfile=/logs/%(program_name)s.err.log [program:log_analyser] command=python -u scripts/log_analyser.py +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) priority=205 stdout_logfile=/logs/%(program_name)s.out.log stderr_logfile=/logs/%(program_name)s.err.log [program:chatbot] command=python -u scripts/chatbot.py +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) priority=220 stdout_logfile=/logs/%(program_name)s.out.log stderr_logfile=/logs/%(program_name)s.err.log -- GitLab