5 years agoWrite correct boolean value for exceede lag, rather than just the lag in seconds. basic_wal_support master
glyn [Thu, 28 Apr 2016 09:39:45 +0000 (10:39 +0100)]
Write correct boolean value for exceede lag, rather than just the lag in seconds.

5 years agoFix scenario where xlog locations differ but apply_lag has returned 0
glyn [Wed, 27 Apr 2016 16:50:25 +0000 (17:50 +0100)]
Fix scenario where xlog locations differ but apply_lag has returned 0

In this case we don't know how far we are behind time-wise, so just set to 1 week;
this is where using a lag value as bytes rather than seconds would be better.

5 years agoAdd some basic support for binary/WAL replicas.
glyn [Wed, 27 Apr 2016 16:09:26 +0000 (17:09 +0100)]
Add some basic support for binary/WAL replicas.

Would be nice to also alter this to support replication slots.

6 years agoAdd support for auth_user and ensure nodes currently being cloned are never used.
glyn [Tue, 13 Oct 2015 09:56:57 +0000 (10:56 +0100)]
Add support for auth_user and ensure nodes currently being cloned are never used.

6 years agoMark experimental smartmach feature as enabled to suppress warnings in perl 5.18+.
glyn [Thu, 6 Aug 2015 12:42:13 +0000 (13:42 +0100)]
Mark experimental smartmach feature as enabled to suppress warnings in perl 5.18+.

This doesn't mitigate the fact that smartmatch is now considered experimental,
but use of smartmatch in this script is limited to simple given/when with string literals.

6 years agoAdd an initial state to cluster hash, this way we can tell the difference between...
glyn [Tue, 13 Jan 2015 13:45:58 +0000 (13:45 +0000)]
Add an initial state to cluster hash, this way we can tell the difference between a cluster that has no suitable target nodes and where the previous cluster state has not yet been initialized.
We also want to register that things have changed and there is now no suitable node, that way we can write nothing and effectively shut off the pool.

6 years agoWhen we repoint an rw server it is perhaps better to force clients to reconnect with...
glyn [Mon, 12 Jan 2015 11:27:47 +0000 (11:27 +0000)]
When we repoint an rw server it is perhaps better to force clients to reconnect with a restart

7 years agoChanges:
glyn [Mon, 1 Dec 2014 20:51:03 +0000 (20:51 +0000)]
- Ensure we always read Slony configuration from the origin node when possible
- Add max_ro_lag=<n> seconds parameter to allow forced reconfiguration of read-only pools when subscribers start to lag

7 years agoAdd in an example template file for a read only slave pool
glyn [Fri, 24 Oct 2014 12:31:05 +0000 (13:31 +0100)]
Add in an example template file for a read only slave pool

7 years agoMinor changes:
glyn [Fri, 24 Oct 2014 12:29:00 +0000 (13:29 +0100)]
Minor changes:
- We should detect changes to server conninfo
- Fixed bug when checking csv sets where we were looking at the wrong value

7 years agoFix regression of commit d6f52cf8280c4f740170bf1f93d31de2e0fd162
glyn [Wed, 22 Oct 2014 13:29:59 +0000 (14:29 +0100)]
Fix regression of commit d6f52cf8280c4f740170bf1f93d31de2e0fd162

7 years agoA couple of minor changes:
glyn [Wed, 22 Oct 2014 13:17:15 +0000 (14:17 +0100)]
A couple of minor changes:

- Added extra configuration parameter 'only_follow_origins' to only repoint pgbouncer configuration if the origin of the monitored sets changes, which is what we usually want.
- Only write cluster status information file when cluster status has actually changed.

7 years agoDon't attempt to fix undefined status file
glyn [Thu, 5 Jun 2014 12:21:11 +0000 (13:21 +0100)]
Don't attempt to fix undefined status file

7 years agoFormat
glyn [Thu, 5 Jun 2014 11:48:37 +0000 (12:48 +0100)]

7 years agoInitial commit
glyn [Thu, 5 Jun 2014 11:35:50 +0000 (12:35 +0100)]
Initial commit