[systemd] decrease start limit and interval

[jbudz]

By default systemd will allow restarting a service 5 times in 10 seconds before entering a failed state. Depending on how Kibana is configured we may not hit this limit, and the service will actively try to restart forever.

This modifies our settings to restart 3 times within a 30 second period before entering a failed state.

Closes #10895

Testing

1. dpkg -i kibana.deb, yum install kibana.rpm etc
2. /etc/kibana/kibana.yml -> invalid setting, e.g. kibana.pid: var/run/kibana.pid (permissions
3. service restarts three times and then stops, can be checked with systemctl status kibana.service

[elasticmachine]

Pinging @elastic/kibana-operations

[mistic]

LGTM

[elasticmachine]

💚 Build Succeeded

• continuous-integration/kibana-ci/pull-request

[bluikko]

These changes should be in the [Service] section and not in the [Unit] section:

systemd[1]: [/etc/systemd/system/kibana.service:3] Unknown lvalue 'StartLimitIntervalSec' in section 'Unit'
systemd[1]: [/etc/systemd/system/kibana.service:4] Unknown lvalue 'StartLimitBurst' in section 'Unit'

On EL7.
It does not start due to the errors and after moving them to [Service] section it works normally.

[jbudz]

Most likely related to systemd versioning, it was moved from the Service section to the Unit section in 229. Can you share your version? /usr/lib/systemd/systemd --version

[jbudz]

I opened #33326

[bluikko]

You are probably correct, the version is 219. Everyone on EL7 should have this same problem, in this version:

• StartLimitBurst is in Service section.
• StartLimitIntervalSec is named StartLimitInterval.

It is very unfortunate that such a critical POS has unstable interfaces on top of other problems.