Skip to content

Commit

Permalink
TODO upgrade and removal of haproxy from the list of tools we use.
Browse files Browse the repository at this point in the history
  • Loading branch information
alexeyklyukin committed Apr 30, 2015
1 parent c841514 commit 41bcbed
Show file tree
Hide file tree
Showing 2 changed files with 24 additions and 6 deletions.
4 changes: 1 addition & 3 deletions postgres-appliance/README
Original file line number Diff line number Diff line change
@@ -1,10 +1,8 @@
Docker image to run the postgres appliance.

It consists of PostgreSQL itself, managed by governor,
etcd proxy and haproxy.
and etcd proxy.

At the moment, etcd is a stand-alone service (and not a proxy),
and haproxy is not used altogether (although installed on the machine).

At the moment, is just a dockerized version of HA template:
https://github.com/compose/governor
Expand Down
26 changes: 23 additions & 3 deletions postgres-appliance/TODO
Original file line number Diff line number Diff line change
@@ -1,17 +1,37 @@
TODO items:

- initialize the appliance with certain parameters,
Done:
initialize the appliance with certain parameters,
like the cluster name and the adddress of etcd.
Use those parameters when configuring the governor,
and etcd-proxy.

- run etcd proxy instead of full-scale etcd
Done:
run etcd proxy instead of full-scale etcd

- configure haproxy to fetch information from
Cancelled, since we are not going to use haproxye:
configure haproxy to fetch information from
the etcd-proxy and point to the acting master

In progress:
- use WAL-E to ship backups and WAL segments to S3

Currently, WAL-E is installed on the Docker image, but not configured.
The way it is planned to be implemented ATM is:

- the governor creates a bucket, which is named by convention, i.e.:
s3://example/acid/postgres/cluster_name/wal

- the AMI is set on the image (currently it's not setting S3FullAccess, TBD to figure out why)
- governor adds WAL-e archive_command to o the recovery conf, i.e
'wal-e --aws-instance-profile --s3-prefix=s3://zalando/acid/postgres/$SCOPE/wal wal-push "%p"'
and to restore_command:
'wal-e --aws-instance-profile --s3-prefix=s3://zalando/acid/postgres/$SCOPE/wal wal-fetch "%f" "%p"'

Additionally, backup policy should be set and regular base backups should be configured in such a way,
that only a certain amount of base backups and WAL segments is stored on AWS (via wal-e delete before)
We might also use the support for backup encryption/compression provided by WAL-E.

- ship PostgreSQL logs somewhere (TBD)

Out of scope for the docker appliance:
Expand Down

0 comments on commit 41bcbed

Please sign in to comment.