Registry downtime 2014-02-16

The npm registry and website were mostly down for about 40 minutes today, from 6.51pm to 7.33pm Pacific Time. The root cause was network issues with our hosting provider, Joyent, possibly related to a DDoS attack on another provider at that same facility.

While nothing went wrong with our systems directly, we can avoid this kind of outage by better distributing our systems across multiple data centers. This was something we were already in the process of doing (the registry never fully went down, for this reason) but given today’s incident we will be accelerating this process.

In particular, we will be mirroring package binaries (current hosted in only one DC) to another location, so that Fastly (our CDN provider) can fall back to them when the main DC is unavailable. This will mean the registry will be available in “read only” mode, even if one datacenter is entirely unavailable.