I have set it up with ECS service, it is able to start, connect to RDS postgres & finish it's initialisation. It registers to a load balancer, which calls /api/health - I can see that it even returned 200 to that.
After that, I find this line in logs after which it goes to shut down.
INFO setting.cache :: e[31mSettings have been changed on another instance, and will be reloaded here.e[0m
Does this make sense? What is it trying to do? How do I get it to not restart?
This is how the log looks in last few lines.
2023-08-25T21:08:37.537+05:30 2023-08-25 15:38:37,537 INFO metabase.task :: Task scheduler initialized into standby mode. Link
2023-08-25T21:08:37.539+05:30 2023-08-25 15:38:37,539 INFO metabase.task :: Initializing task SyncDatabases 📆 Link
2023-08-25T21:08:37.760+05:30 2023-08-25 15:38:37,760 INFO task.sync-databases :: Updated default schedules for 0 databases Link
2023-08-25T21:08:37.761+05:30 2023-08-25 15:38:37,761 INFO metabase.task :: Initializing task PersistRefresh 📆 Link
2023-08-25T21:08:37.862+05:30 2023-08-25 15:38:37,862 INFO driver.impl :: Initializing driver :sql... Link
2023-08-25T21:08:37.863+05:30 2023-08-25 15:38:37,863 INFO driver.impl :: Initializing driver :sql-jdbc... Link
2023-08-25T21:08:37.864+05:30 2023-08-25 15:38:37,864 INFO driver.impl :: Initializing driver :h2... Link
2023-08-25T21:08:38.055+05:30 2023-08-25 15:38:38,055 INFO metabase.task :: Initializing task CheckForNewVersions 📆 Link
2023-08-25T21:08:38.264+05:30 2023-08-25 15:38:38,264 INFO metabase.task :: Initializing task PersistPrune 📆 Link
2023-08-25T21:08:38.342+05:30 2023-08-25 15:38:38,342 INFO metabase.task :: Initializing task SendAnonymousUsageStats 📆 Link
2023-08-25T21:08:38.376+05:30 2023-08-25 15:38:38,375 INFO metabase.task :: Initializing task ModelIndexValues 📆 Link
2023-08-25T21:08:38.382+05:30 2023-08-25 15:38:38,381 INFO metabase.task :: Initializing task RefreshSlackChannelsAndUsers 📆 Link
2023-08-25T21:08:38.551+05:30 2023-08-25 15:38:38,551 INFO metabase.task :: Initializing task SendPulses 📆 Link
2023-08-25T21:08:38.578+05:30 2023-08-25 15:38:38,578 INFO metabase.task :: Initializing task SendFollowUpEmails 📆 Link
2023-08-25T21:08:38.644+05:30 2023-08-25 15:38:38,643 INFO metabase.task :: Initializing task TaskHistoryCleanup 📆 Link
2023-08-25T21:08:38.676+05:30 2023-08-25 15:38:38,676 INFO metabase.task :: Initializing task TruncateAuditLog 📆 Link
2023-08-25T21:08:38.787+05:30 2023-08-25 15:38:38,786 INFO core.QuartzScheduler :: Scheduler MetabaseScheduler_$_ip-172-20-3-213.ec2.internal1692977916437 started. Link
2023-08-25T21:08:38.789+05:30 2023-08-25 15:38:38,788 INFO metabase.task :: Task scheduler started Link
2023-08-25T21:08:38.790+05:30 2023-08-25 15:38:38,790 INFO metabase.core :: Metabase Initialization COMPLETE in 2.0 mins Link
2023-08-25T21:08:38.952+05:30 2023-08-25 15:38:38,951 INFO task.refresh-slack-channel-user-cache :: Slack is not configured, not refreshing slack user/channel cache. Link
2023-08-25T21:08:47.722+05:30 2023-08-25 15:38:47,721 INFO setting.cache :: Settings have been changed on another instance, and will be reloaded here. Link
2023-08-25T21:08:47.736+05:30 2023-08-25 15:38:47,735 INFO metabase.core :: Metabase Shutting Down ... Link
2023-08-25T21:08:47.737+05:30 2023-08-25 15:38:47,736 INFO core.QuartzScheduler :: Scheduler MetabaseScheduler_$_ip-172-20-2-137.ec2.internal1692977589168 shutting down. Link
2023-08-25T21:08:47.737+05:30 2023-08-25 15:38:47,737 INFO core.QuartzScheduler :: Scheduler MetabaseScheduler_$_ip-172-20-2-137.ec2.internal1692977589168 paused. Link
2023-08-25T21:08:47.756+05:30 2023-08-25 15:38:47,747 INFO core.QuartzScheduler :: Scheduler MetabaseScheduler_$_ip-172-20-2-137.ec2.internal1692977589168 shutdown complete. Link
2023-08-25T21:08:47.757+05:30 2023-08-25 15:38:47,757 INFO metabase.server :: Shutting Down Embedded Jetty Webserver Link
2023-08-25T21:08:47.772+05:30 2023-08-25 15:38:47,771 INFO metabase.core :: Metabase Shutdown COMPLETE```Okay, figured the shutdown is from outside factor. CPU wasn't enough. There are certain examples out there on the internet, which are showing 0.5 CPU is enough - which apparently is not.
Making it to 2 vcpu works.