Running cluster replication over flaky connection


(Mikhail Sayapin) #1

Hello everyone,

I have a PC at home in China, which is downloading some data from Chinese
websites and processing it for further use. On the same PC there is an ES
master node, which indexes these data.

Another server in Europe is holding the copy of the index, and used only
for reading.

So basically Chinese server is Master, and Europe server is slave.

The connection is very flaky and slow, bandwidth (256 kbit, yes) is ok, but
sometimes network splits, and the packet loss is about 5%.

Could you please advise what network and cluster settings I should tweak to
ensure seamless operation?

I'm thinking of setting "transport.tcp.compress" to true, of course. Also I
believe I should set data: true for the Europe node. Beyong that I'm a bit
lost and would appreciate any input.

Thanks!

--


(system) #2