I work at a company where we have a cluster of 4 machines

with PG and Slony-I. These 4 machines are split in two data

centers, so we have one ´master´ and one ´slave´ on each data

center, and then one of the masters is the slave of the other:



   DC1        DC2

  Master --> Master

    |          |

    V          V

  Slave      Slave



  Our setup ensures that writes to the DB (which are done

only a few times a day by big batches with) are only done on

THE master (in the graph, the master in DC1).



  I have three questions about this setup:



  What happens if THE master dies? We can stop the batches

that inject the data, but will the slaves hold the reads

(several thousand a day) while we re-establish THE master?

And what exactly happens when THE master is back up and we

restart the batches?



  What if a slave is out for a day (for instance, HW testing)?

Can we just read it to the cluster? The current internal

procedure involves dropping the node, dropping the slony

schema, rebuilding the node and resuscribing the replication

lines, which includes a full data copy.



  Which would be a good way to invert the relationship between

the masters? The internal procedure involves stopping the

batches, changing the subscriptions and restarting the batches

(during the night, when the activity is low).
_______________________________________________
Slony1-general mailing list
[email protected]
http://lists.slony.info/mailman/listinfo/slony1-general

Reply via email to