tracking the highest observed timeline in the cluster

[YanChii]

Hi,
as I promised in #57, I've implemented the new data security check.

Now it is probably time to discuss one issue. For tracking the highest reached timeline, I've created a unique variable name that includes a postgresql DB system identifier in its name (so the name of the attribute is e.g. "postgres-6354064916801227111-highest-timeline").
I did this because:
1), the minor issue: if you delete all datadirs and re-init the whole cluster, you start from timeline 1 again and we don't use the old parameter.
2), the major issue: if you have two databases with two configured HA resources, they will interfere as they use the same parameter names.

An the second issue is what I'm concerned for also with the other parameters you already have.
My question:
If you have two pgsqlms resources within one pacemaker cluster, can attributes (lsn_location, nodes, cancel_switchover) interfere between them? For example, if master node fails and both resources start master voting simultaneously, will they write to the same attributes and therefore master voting can work with wrong data?

Thank you for your consideration & reply.

Jan

[ioguix]

Answered in #57

[ioguix]

I pushed another patch in master in regard with the TL check during standby election.

See b75971d

I'll close this pull request for now. Discussion following on #57