-
Type: Improvement
-
Resolution: Duplicate
-
None
-
Affects Version/s: None
-
Component/s: None
-
Environment:Ubuntu 12.04 (30+ servers)
*Location*: http://docs.mongodb.org/manual/reference/program/mongos.exe/
*User-Agent*: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/27.0.1453.110 Safari/537.36
*Referrer*: http://docs.mongodb.org/manual/search/?query=mongos
*Screen Resolution*: 1680 x 1050
*repo*: docs
*source*: reference/program/mongos.exe
Ubuntu 12.04 (30+ servers) *Location*: http://docs.mongodb.org/manual/reference/program/mongos.exe/ *User-Agent*: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/27.0.1453.110 Safari/537.36 *Referrer*: http://docs.mongodb.org/manual/search/?query=mongos *Screen Resolution*: 1680 x 1050 *repo*: docs *source*: reference/program/mongos.exe
Hi there,
We are deploying a potentially very large mongo based infrastructure. I recentlky had casue to move from one server supplier to another. Luckily I was using host names. In one area we have 3 mongos units working over a sharded cluster.
What I found strange was that to move the mongos units we had to copy the actual data files over, which we did successfully with zero downtime btw. However, surely the mongos units should be treated as a simple replica set? Tis way we could start up new units, make them a part of teh repl set, and then close down teh old ones.. as you would with a normoal mongod replica set?
Just a thought
Peter Colclough
peter@madbits.com
Reporter: Peter Colclough
E-mail: peter@madbits.com