garage_net: retry connecting when new IP is learned
Changing IP address of a node leads to a half-connected and broken cluster
3-degraded
mode with retry to avoid 404 not found
This is actually quite dangerous, because we could then read an inconsistent value from the versions table, e.g. a version with only a subset of the actual data blocks. And in that case we don't…
garage_net: retry connecting when new IP is learned
Add a local lock on CreateBucket etc, as a temporary measure
Add config option for bind ip for outgoing connections
When one node is running behind NAT, the cluster gets confused with wrong rpc_public_addr, not respecting setting file
doc: fixes to index of configuration options