Hi Farshid & team,
Two things pop to mind...
First, getting the output of a "curl http://HOST:8091/pools/default/buckets
" would be a useful next step. What seems to be happening is the cmd-line parameter of "http://192.168.42.16:8091
" says to try to use the server "192.168.42.16". However, that (poorly written) error message means that "192.168.42.16" is not appearing in the REST/JSON server map, so the cbbackup errors out saying there isn't a good source-node to use to backup the design-doc.
The workaround is for the user to specify the same server IP/hostname that appears in the server-map, and that should work.
Secondly, the "-t" threads parameter of 128 is probably quite too high and might steal from performancel. If there are 4 nodes in the cluster, for example, even with -t 128, only 4 threads of those 128 worker threads would have any work to do.