- Sep 30, 2019
-
-
- Sep 11, 2019
-
-
Lawrence Jones authored
Reusing FindDB method and added test for the same
-
- Aug 29, 2019
- Jul 31, 2019
-
-
Simone Gotti authored
Release v0.14.0
-
- Jul 26, 2019
-
-
Simone Gotti authored
-
Simone Gotti authored
-
Simone Gotti authored
Update architecture.md
-
- Jul 20, 2019
-
-
Simone Gotti authored
Fix hanging sentinels
-
- Jul 19, 2019
-
-
Yujun Zhang authored
Fix typo
-
- Jul 16, 2019
-
-
Lawrence Jones authored
If etcd returned an error, our previous behaviour would be to report the error and restart an election. The outer loop never responded to the error by stopping the election, which would cancel the context, and instead would restart a new campaign. The first campaigns session will continue to exist on the context the election loop had originally provided it, which is context.Background. We'll come round to perform a new leadership election and the etcdElection.Campaign() method will block waiting for all existing leased keys to be removed from the prefix. Because the old campaign sessions are still alive, they'll continue to renew their leases and ensure those revisions never get deleted, thus blocking our sentinel campaign loop indefinitely. After discussion in #659, it became clear that most of the store implementations are likely to have similar issues, and instead of fixing the underlying implementations we decided to amend the interface to warn consumers that they need to terminate existing elections before starting new ones. This is a sharp edge of the election API and one that might catch us in a future refactor. This is balanced against a hope that we'll refactor the election interface before a second use case is likely to arise.
-
- Jul 03, 2019
-
-
Simone Gotti authored
Enable verbose & progress logging for pg_basebackup
-
- Jun 18, 2019
-
-
Harry Maclean authored
This will cause pg_basebackup to output some extra steps during startup and shutdown. If progress reporting is also enabled (-P) then it will show the exact file name that is currently being processed. This can be useful for debugging and monitoring the backup process. Combined with verbose mode (-v) this will give detailed information on what files are being processed and how far through the backup we are. There's a slight initial time cost introduced as Postgres has to determine the total database size, but this should be minimal compared to the backup time. Co-authored-by: <me@lawrencejones.dev> Co-authored-by: <harrymaclean@gmail.com>
-
- Jun 13, 2019
-
-
Simone Gotti authored
tests: fix missing rebase of PR #668
-
Simone Gotti authored
My fault since I disabled github option to require branches to be up to date before merging to avoid asking people to always rebase after every PR merge...
-
Simone Gotti authored
integration tests: improve reliability
-
Simone Gotti authored
Label cluster identifier with component
-
Simone Gotti authored
* Increase TestAutomaticPgRestart timeouts since they are too short. * Log stolonctl output so if it fails we can know the reason. On semaphore: * Reduce parallel test to 5. This will make tests take almost 50 minutes... * Use libeatmydata to skip postgres/initdb fsync without changing the various initdb options or postgres parameters. This doesn't work for etcd/consul since go uses a direct syscall instead of libc functions. * Stop some unneeded services that use a lot of memory.
-
- Jun 10, 2019
-
-
Lawrence Jones authored
[^1]: https://github.com/gocardless/stolon-pgbouncer/pull/46 The stolon_cluster_identifier metric is universal to all stolon components, and can be used to join metric series across a stolon cluster. By adding the component label, it's possible to write polymorphic alerts that expose the component name when firing. This change also registers the metric in the sentinel and proxy, as it should be present in all long-lived binaries. This accompanies a change in stolon-pgbouncer to do the same.
-
- Jun 07, 2019
-
-
Simone Gotti authored
keeper: configurable wal_keep_segments
-
- Jun 06, 2019
-
-
Simone Gotti authored
Add badges in README
-
Simone Gotti authored
keeper: fix some logging
-
Simone Gotti authored
Add Timeout for kubeclient
-
Dinesh authored
-
Simone Gotti authored
use the value provided in the pg parameters to set the instance wal_keep_segments value. The value is ignored if it's a bad value or lesser than the default value (8).
-
Simone Gotti authored
Update etcd dependency
-
Dinesh authored
-
- Jun 05, 2019
-
-
Simone Gotti authored
Small dead code removal in Sentinel.go
-
Lawrence Jones authored
- github.com/coreos/etcd v3.2.11+incompatible + github.com/coreos/etcd v3.3.13+incompatible Between v3.2.11 and v3.3.13 there have been a number of changes to the way the etcd client handles failed server connections, notably a significant refactor of the client balancer which appear to have fixed several issues when communicating with etcd through a TCP L4 load balancer.
-
Simone Gotti authored
Remove capital first letter and use zap.Error to wrap errors
-
- Jun 04, 2019
-
-
Simone Gotti authored
Allow special characters in pg-su-username
-
Dinesh authored
-
Dinesh authored
-
- Jun 03, 2019
-
-
Simone Gotti authored
sentinel: Add Prometheus metrics
-
Ben Wheatley authored
These metrics provide the ability to build alerts that tell us whether the sentinels are operating as expected: - The last time that the sentinel successfully processed the clusterdata. - Whether the sentinel is a leader. - The number of times that the sentinel has been elected leader.
-
Simone Gotti authored
Store only last 2 postgres timeline histories
-
- May 31, 2019
-
-
Dinesh authored
-
- May 28, 2019
-
-
Dinesh authored
Co-authored-by: Arunvel Sriram <arunvelsriram@gmail.com>
-
Simone Gotti authored
*: update to go1.12
-
- May 27, 2019
-
-
Simone Gotti authored
Update travis and use external go vet shadow tools since it's not included in the go vet command
-