0

I have 10 servers and want to use all as storage so tried to create volume with following command but its throwing error

gluster volume create gv0 replica 3 server1:/data/brick1/gv0 server2:/data/brick1/gv0 server3:/data/brick1/gv0 server4:/data/brick1/gv0 server5:/data/brick1/gv0 server6:/data/brick1/gv0 server7:/data/brick1/gv0 server8:/data/brick1/gv0 server9:/data/brick1/gv0 server10:/data/brick1/gv0

and it throws error

Numbers of bricks is not a multiple of replica count

what's recommendation in this scenario?

it look like if i create with replica 2 it will succeed, so what would be difference in my scenario if i create volume with replica 2 or replica 3(i am assuming i will use 9 bricks here) on 10 brick?

In replica 2 with 10 brick scenario will i face split brain issue?

ImranRazaKhan
  • 150
  • 3
  • 15

1 Answers1

2

To store 3 copies of everything, you need 3N disks. 10 is not a multiple of 3.

3 copies seems like overkill, especially if they are all in the same building -- that is subject to the same fire, flood, earthquake, hurricane, or whatever. Or can you spread them out geographically?

Rick James
  • 2,463
  • 1
  • 6
  • 13
  • 1
    What would be the impact if i create 9 bricks with replica 3? We have 10 servers and chances are we may shutdown multiple servers(randomly) on base of requirement and any 5 or 6 will be up at any time. – ImranRazaKhan Nov 11 '20 at 19:20
  • 1
    If you will be manually shutting down random servers, the problem _and_ solution is in manual rules and discipline. What if they randomly shutdown all 3 servers in a replica set? You would be hosed. The rule has to be "Never shutdown more than N-1 servers at the same time. (N=3, the replica number.) – Rick James Nov 11 '20 at 19:26
  • if i add 10 servers as peer and create volume with mentioning 3 servers, on shutting of one of server replica set , will gluster include other available server automatically in replicaset? – ImranRazaKhan Nov 11 '20 at 19:32
  • If it must rebuild the data on that new disk, that could take time -- possibly longer than is acceptable. Perhaps the Question should include more on why you are looking at this kind of redundancy -- low reliability? flaky network? other things? There may be a different solution for the real goal. – Rick James Nov 11 '20 at 19:42
  • This is an IOT based solution which connected to particular devices which collect data from solar/wind/grid, So we powerup machines as per requirement when we use solar then solar related servers are up and likewise. Sometime all could be up. – ImranRazaKhan Nov 11 '20 at 19:46