Node.js script tries to run mongodump (nonexistent in app container)

[eecavanna]

The web server application tries to run mongodump.

nmdc-edge/webapp/server/crons/dbBackup.js

Line 11 in 7bcee5b

const cmd = `mongodump --db ${config.DATABASE.NAME} --out ${config.DATABASE.BACKUP_DIR}/db-backup_${dateStringWithTime}`;

The container in which the web app is running does not include mongodump.

Here's a snippet from the logs of the app container on Jetstream2:

// $ docker compose logs app | grep 'mongodump'

exouser-app-1  | 2024-07-01 01:00:00 info: 	mongodump --db nmdcedge --out /project/io/db/db-backup_2024-07-01:01:00
exouser-app-1  | 2024-07-01 01:00:00 error: 	Command failed: mongodump --db nmdcedge --out /project/io/db/db-backup_2024-07-01:01:00
exouser-app-1  | /bin/sh: mongodump: not found
exouser-app-1  | 2024-07-01 01:00:00 error: 	/bin/sh: mongodump: not found

[eecavanna]

Looks to me like that command dumps the nmdcedge database to a directory (within some fixed base directory) named YYYY-MM-DD:HH:mm (e.g. db-backup_2024-12-25:12:59).

One concern I have about that is that having colons in a filename may lead to issues in some operating systems (for example, I don't think it's valid in Windows). I'd recommend using underscores, hyphens, or nothing; e.g. db-backup_20241225_1259.

[eecavanna]

The file webapp/server/cronserver.js has the following two jobs (among others) defined in it:

nmdc-edge/webapp/server/cronserver.js

Lines 73 to 80 in bdf4e33

	//backup nmdcedge DB every day at 10pm
	cron.schedule(config.CRON.SCHEDULES.DATABASE_BACKUP_CREATOR, function () {
	dbBackup();
	});
	//delete older DB backups every day at 12am
	cron.schedule(config.CRON.SCHEDULES.DATABASE_BACKUP_PRUNER, function () {
	dbBackupClean();
	});

The server's config.js file has these default schedules for those jobs:

nmdc-edge/webapp/server/config.js

Lines 113 to 114 in bdf4e33

	DATABASE_BACKUP_CREATOR: process.env.CRON_DATABASE_BACKUP_CREATOR_SCHEDULE || "0 1 * * *",
	DATABASE_BACKUP_PRUNER: process.env.CRON_DATABASE_BACKUP_PRUNER_SCHEDULE || "0 2 * * *",

In English, those schedules are:

Job	cron	English
Create	0 1 * * *	At 01:00
Prune	0 2 * * *	At 02:00

[eecavanna]

For the mongo container, we could create a file named schedule_backups_cronjob.sh in /docker-entrypoint-initdb.d and populate it with the following:

# Install dependencies.
apt update
apt install -y cron nano

# Add the job description to the cron table.
echo '0 1 * * *    mongodump --authenticationDatabase admin --username' "${MONGO_INITDB_ROOT_USERNAME}" '--password' "${MONGO_INITDB_ROOT_PASSWORD}" '--db nmdcedge --out "/tmp/db-backup_`date +\%Y\%m\%d_\%H\%M`" >> /tmp/backups_cronjob.log 2>&1' | crontab -

# Restart the cron service (is this necessary?)
service cron start

Replace /tmp with the path to the directory in which we want the backups to be created.

The container will execute that script upon its startup, according to the "Initializing a fresh instance" section of the container image docs.

[ssarrafan]

@eecavanna who is this assigned to? Is it being worked on this week?

[eecavanna]

Hi @ssarrafan, it isn't assigned to anyone currently. I'll assign it to myself. I don't expect to work on it any more this week. I'll move it to the next sprint, as I do plan to continue working on it again within the next couple weeks. At some point (depending upon what the remainder of my digging uncovers), I may reassign it or spin part of it off into a separate ticket.

[ssarrafan]

@eecavanna would you be ok with me moving anything not Berkeley refactor to the next sprint? Like this issues, seems like this could wait.

[eecavanna]

Hi @ssarrafan, yes on this particular issue. I'll move it now.

There are things I want to do that aren't Berkeley Schema Roll Out-related, next Wednesday-Friday.