| Table of Contents |
|---|
...
When to include domain and Host
The only time domain is required is for an SCSP method on a domain object itself. Neither domain nor Host is required for requests within the default cluster domain; otherwise, the domain name must be passed as the Host in the request. (Your cluster should have one domain with the same name as the cluster, which sets up a default cluster domain.)
Client applications most often send the domain name as the Host in the request. When the Host header does not match the domain name, the client can supply the domain argument to explicitly override any value from the Host request header. A domain argument always has precedence over the Host header in the HTTP/1.1 request.
...
| Code Block | ||
|---|---|---|
| ||
METHOD /bucketname/objectname[?query-arguments] HTTP/1.1 |
where
- bucketname is a simple, URL-encoded identifier that cannot contain slash characters (or any other character not allowed in HTTP URLs)
- objectname is any legitimate URL, which can contain slash characters
...
You specify the UUID with all SCSP methods except WRITE, in which case the cluster will return the UUID in the response if the write is successful.
| Info | ||
|---|---|---|
| ||
| When writing unnamed objects, use a HOST header equivalent to the cluster name, the host IP address, or a domain=clusterName query arg on all requests even if you are not using domains for other purposes. |
...
| Info | ||
|---|---|---|
| ||
| When writing unnamed objects, ensure that your application is not passing a HOST header that is neither an IP address nor a domain that exists in the cluster (unless the host header matches the cluster name). Swarm will attempt to look up the non-existent domain on every request and will wait for multiple retries before the lookup times out, impacting performance. |
...