tree 0d4ab65ff6af76e3115ac8a82452216301aebce4
parent 72068da814af80568cb106b877ef8f5e526e684c
author Serge Bazanski <serge@nexantic.com> 1615897053 +0100
committer Serge Bazanski <serge@nexantic.com> 1615897053 +0100

m/node/core/conensus: handle etcd restarts

This makes the etcd service more reliable in case of transient failures
when starting in a new cluster. Previously, any restart of etcd on the
first (bootstrapping) node would cause etcd to get stuck and never start
again (as certificates were already created). This changes the logic to
allow existing certificates.

This also handles the case of etcd attempting to start as the network is
reconfigured, and eg. the external hostname is not yet resolvable.

Test Plan:
No tests yet. This should be tested by a more comprehensive e2e test where we
randomly kill some runnables (see: T872).

X-Origin-Diff: phab/D733
GitOrigin-RevId: 8ac426f9423ec2353537eec651071e99a5e5ec53