agreed there is an element of risk.
however a backup of the master will be carried out prior (ok there could potentially be a gap of data during downtime)
unfortunately i have nothing else big enough (diskwise) to run my tests on.
On 30 June 2015 at 15:47, Michael Paquier <michael.paquier@xxxxxxxxx> wrote:
On Tue, Jun 30, 2015 at 2:29 PM, Andy Erskine <andy.erskine@xxxxxxxxxx> wrote:Ok so a question i should have asked at the very beginning ..If i touch my trigger file - promoting the secondary to a master - will that in anyway effect the master thats already running ?IE no files on the master will change ?Streaming replication is designed to let the master run properly when a standby disconnects.
Then all i'll have to do is shutdown the secondary when i've finished and run the basebackup process again to restore replication (and change recovery.done - conf again.--
Yes.
Now something that has been mentioned by the others: isn't your standby here for a reason? Like in case of failure don't you have a process to failover automatically? Perhaps you are shooting yourself in the foot by unplugging this standby, hence you should, and other recommend the same, simply let the existing standby alone and create a new instance by taking a new base backup from either the master or the standby and use it for your tests. Then eliminate the node you created. In short: avoid doing stupid things...Michael
Andy Erskine
JDS Australia
P 1300 780 432 | M 0431919301 | F 03 90124578
E: andy.erskine@jds.net.au
W: www.jds.net.au
Level 8, 2 Russell Street, Melbourne, VIC 3000
GPO Box 4777, Melbourne VIC 3001
