Improvement/Provide a cron that upgrades the hub and the collections/parsers/scenarios/....

[blotus]

[cmiscloni]

+1

[okohll]

Does

#!/bin/bash
cscli hub update
cscli hub upgrade
cscli collections upgrade -a
cscli parsers upgrade --all
cscli scenarios upgrade -a
systemctl reload crowdsec

about cover everything?

[LaurenceJJones]

#1817 Fixed this

[mmetc]

There are still a few issues

• crowdsec does not reload the configuration after hub upgrade
• the package manager doesn't know about the cron job, so it won't be able to update it, or see if the user made changes
• wizard.sh is soon going be a /bin/sh script - not bash - for compatibility with other distributions. It won't be called nor sourced from the postinst file. Also passing arguments to sourced files (the -n for noop) only works with bash

For these reasons, it would be better to follow each distribution's guidelines and install the cron job like we would install any other file:

• https://fedoraproject.org/wiki/Packaging:CronFiles
• https://www.debian.org/doc/manuals/maint-guide/dother.en.html#crond

A minimal cron.daily/crowdsec-hub would look like:

#!/bin/sh
cscli --error hub update && cscli --error hub upgrade
systemctl reload crowdsec
exit 0

I'm not convinced that systems installed with wizard.sh should have cron jobs, but if they do, they might not have systemd. That part is WIP.

What do you think? @sabban ?

[mridullpandey]

hi @mmetc i want to take this up..

[mmetc]

Hi @mridullpandey

It's been merged here https://github.com/crowdsecurity/crowdsec/blob/master/config/crowdsec.cron.daily

Any suggestion? We're releasing soon

Thanks

[tasiotas]

what about docker image? could you include this cron in docker as well?

[mmetc]

@tasiotas there are several drawbacks to running a process manager and cron in a container, we'll try and find a better option for both docker and k8s

[jflattery]

@tasiotas there are several drawbacks to running a process manager and cron in a container, we'll try and find a better option for both docker and k8s

Is there a issue that can be tracked for this? Is there a "proper" way to do this as of today?