brick_squorum module should return more informative quorum_error response

hibari / gdss-admin

http://hibari.github.com/hibari-doc/

Other

2 stars 0 forks source link

brick_squorum module should return more informative quorum_error response #9

Open tatsuya6502 opened 9 years ago

tatsuya6502 commented 9 years ago

Current error response from brick_squorum module is just an atom quorum_error. Need more info for troubleshooting.

RPC(bootstrap) to hibari@hibari1 failed: {'EXIT',
                                          {{{badmatch,quorum_error},
                                            [{brick_admin,
                                              bootstrap_existing_schema2,1,
                                              [{file,"src/brick_admin.erl"},
                                               {line,999}]},

It would be nice if an error response contains the following info.

response mismatch
response timestamp mismatch
not enough number of bricks available

Sacrifiser commented 7 years ago

2 ADMIN_NODES is not supported

Can some one write how to fix this issue ? 0.1.11

tatsuya6502 commented 7 years ago

2 ADMIN_NODES is not supported

@Sacrifiser Thanks for your comment. 0.1.11 does support 2 admin nodes configuration but this configuration is not recommended for production system. Because, in some failure scenarios, we found admin servers may get stale information about the chain configuration and can go into split-brain situation.

We recommend 3 admin nodes configuration, which can only tolerate 1 node down at a time, but will avoid the split-brain situation. If you have only 2 machines for storage nodes, consider adding 1 small machine for only running third admin node.