etcd: Update to v3.4.10

[surajssd]

Release Notes: https://github.com/etcd-io/etcd/releases/tag/v3.4.10

[invidian]

There is a breaking change for existing clusters: https://github.com/etcd-io/etcd/blob/master/CHANGELOG-3.4.md#breaking-changes.

[surajssd]

@invidian the breaking change was fixed.

[surajssd]

I'd like to understand what was the breaking change and how it was fixed.

See the last commit's description. But tl;dr is that etcd 3.4.10 would fail if it found some other permissions on the data dir. But with this change the permission of the dir will be 0700.

Also I have verified with lokoctl this does not trigger cluster recreation. The upgrade process has to be manual on the host.

---

Run lokoctl cluster apply with this PR commit. Now ssh into the controller node and run following commands:

# Become root
sudo -i

# Change the permissions as required for the newest version
chmod 0700 /var/lib/etcd/

# Verify the dir permissions
ls -ld /var/lib/etcd/

# Update the version
sed -i 's|3.4.9|3.4.10|g' /etc/systemd/system/etcd-member.service.d/40-etcd-cluster.conf

# Restart the process and verify that it is in Running state
systemctl daemon-reload
systemctl restart etcd-member
systemctl status etcd-member


# Verify as a user:
cd /opt/bootkube/assets/tls/
export ETCDCTL_API=3
export endpoints=https://foobar.net:2379
export cacert=etcd-client-ca.crt
export cert=etcd-client.crt
export key=etcd-client.key
etcdctl member list --cacert=$cacert --cert=$cert --key=$key --endpoints=$endpoints

Can you verify this works @ipochi?

[surajssd]

Created a doc which explains how to upgrade the etcd cluster #802.

[surajssd]

@rata @ipochi PTAL.

[rata]

LGTM, thanks!

[invidian]

OK

[ipochi]

LGTM