Replication and failover strategy for servers

Question

Replication and failover strategy for servers

shumkov opened this issue 13 years ago · 4 comments

shumkov commented 13 years ago

Answer 1 · 2012-05-12T15:58:25.000Z

I wonder if this shouldn't be handled in code but by a TCP loadbalancer like haproxy?

Answer 2 · 2012-05-17T13:37:19.000Z

No, its 'readOnly', 'writeOnly' options for servers

Answer 3 · 2015-11-30T19:51:50.000Z

What happens currently if 2+ servers are configured with consistent distribution hashing (ketama) but one of the servers becomes unavailable?

Answer 4 · 2015-12-03T13:04:57.000Z

According to my research and tests, there is zero special handling for server failure. The consistent key hashing will distribute cache data across a pool of servers. However, if a member of the pool is unavailable, and the key you're operating on hashes to that server, you have a fatal situation for that particular key.