[Pgpool-general] Online Recovery/Dynamic Addition of Node

[Harold Lim]

Hi All,

Is it correct to say that in PGpool, if the current running workload is heavy, the time it also take to finish online recovery or dynamic addition of node will also be longer?

Is this a normal behavior or is there something wrong with how I set up pgpool/postgresql:

I want to test and see the behavior/performance of pgpool when it dynamically adds a new node.

I have set up 1 back end node and while pgbench (I set the number of transactions to be around 44000) is running, I start the online recovery process. The problem is it takes a really long time. It doesn't even complete until after pgbench finishes (20+ minutes). 
In my postgresql log I see something like this: 
tar: pg_xlog/000000170000000000000045: file changed as we read it


Without running any workload, recovery takes only a few minutes to complete.

Is this behavior duty to the fact that to perform online recovery, in the first stage, it has to wait for all of the clients to disconnect first?

Thanks,
Harold