Alexander Kukushkin
a1f11fe2fe
Remove unneeded class CurrentLeaderError
2015-07-08 08:53:08 +02:00
Alexander Kukushkin
43b12af3a7
Implement possibility to work against ZooKeeper
...
This implementation is using the same interface (AbstractDCS) as Etcd
class. It means that there should be no problem to implement another
plugin to work agains Consul for example.
2015-07-07 12:45:14 +02:00
Alexander Kukushkin
7445590546
Keep node name in AbstractDCS class
...
It eliminates need to pass this name into most of the methods
2015-07-07 12:26:52 +02:00
Alexander Kukushkin
c5edcb589b
return result of last_operation as a string
2015-07-07 12:24:02 +02:00
Alexander Kukushkin
1531304baa
return result of last_operation as a string
2015-07-07 12:23:06 +02:00
Alexander Kukushkin
ae689a2f83
Rename ha.etcd into ha.dcs
2015-07-02 15:33:21 +02:00
Alexander Kukushkin
cc71906009
Inherit Etcd from abstract class
2015-07-02 15:15:07 +02:00
Alexander Kukushkin
6079c8359f
refactoring: preparing to support ZooKeeper
...
move some common classes info separate file in preparation to support distributed configuration store other then Etcd
2015-07-02 12:06:07 +02:00
Alexander Kukushkin
687fb7da10
Remove haproxy_status.sh and change haproxy.cfg to work directly with the governor
2015-06-25 14:17:51 +02:00
Alexander Kukushkin
a723dbbed4
Merge branch 'master' of github.com:compose/governor
2015-06-25 14:15:52 +02:00
Alexander Kukushkin
9732106e5d
BUGFIX: keep list of replication slots on slaves empty
...
Otherwise slaves accumulate wal files...
2015-06-25 14:01:06 +02:00
Alexander Kukushkin
a02f9d497f
BUGFIX: it can be that /members request to etcd returns somesing strange with status_code 200
2015-06-24 16:41:37 +02:00
Alexander Kukushkin
9fc0fa9f57
Merge pull request #4 from CyberDem0n/etcd-cluster-support
...
real_ttl method will return -1 when expiration is not defined
2015-06-23 17:16:26 +02:00
Alexander Kukushkin
f7be52f63a
real_ttl method will return -1 when expiration is not defined
2015-06-23 17:15:37 +02:00
Alexander Kukushkin
2e62f4a2bf
Merge pull request #3 from CyberDem0n/etcd-cluster-support
...
Etcd cluster support
2015-06-23 14:31:12 +02:00
Alexander Kukushkin
b7c73fdba8
Rename hostname to name in a Member
2015-06-23 14:07:00 +02:00
Alexander Kukushkin
744026b4bb
Do not update member TTL when it is far from being expired
2015-06-23 14:02:32 +02:00
Christopher Winslett
4871a25742
typo in Readme
2015-06-22 14:15:40 -07:00
Alexander Kukushkin
44e06b4cff
Add expiration field into Member
2015-06-10 14:22:11 +02:00
Alexander Kukushkin
275e18e58a
Refactor tests
2015-06-10 14:21:23 +02:00
Alexander Kukushkin
971bc3b6e8
Increase verbosity of failed PUT and DELETE requests. Hope will help to debug some etcd related issues
2015-06-10 11:56:16 +02:00
Alexander Kukushkin
a64dad267e
Bugfix: target field of SRV object should be casted to string
2015-06-09 15:46:34 +02:00
Alexander Kukushkin
57914821d0
Randomize list of etcd cluster members
2015-06-09 12:30:58 +02:00
Alexander Kukushkin
a2988361c5
Retry with host from config when it failed to resolve SRV
2015-06-09 12:30:16 +02:00
Alexander Kukushkin
261fbf2721
Replace logging with logger in ha.py
2015-06-09 10:27:09 +02:00
Alexander Kukushkin
da4f9e8445
Make logging of http requests more verbose
2015-06-09 10:25:33 +02:00
Alexander Kukushkin
5b7a60657f
Fix syntax in .travis.yml
2015-06-09 10:04:07 +02:00
Alexander Kukushkin
c79fba7656
Support work with etcd as a cluster
...
In case if one member of a cluster is not available it will retry with
another one and fetch the new cluster configuration. Default timeout for
all requests to etcd is 5 seconds.
Initial cluster configuration can be resolved through:
1) /v2/members call on one of the cluster members on a client port
2) when it is possible to resolve hostname into multiple ip's it will
iterate through list and try to perform action from 1)
3) If there is discovery_srv defined in etcd section of config file it
will resolve peer addresses of all cluster members and will fetch
cluster configuration with using peer protocol by doing /members call on
a peer port
2015-06-09 09:51:53 +02:00
Alexander Kukushkin
fad2fbeae0
Import ThreadingMixIn for python3
2015-06-03 16:39:29 +02:00
Alexander Kukushkin
aa1fc315ae
Process http requests in a threads with ThreadingMixIn
2015-06-03 16:22:05 +02:00
Alexander Kukushkin
f9fbe8980f
Implement lsn_to_bytes bytes_to_lsn functions
2015-06-02 13:36:36 +02:00
Alexander Kukushkin
d9fbbf3ae9
Make pg_hba section more compact
2015-06-02 12:37:26 +02:00
Alexander Kukushkin
fca987deda
set value of api_url from application_name parameter of conn_url. Reset parameters of conn_url
2015-06-02 09:45:37 +02:00
Alexander Kukushkin
f9ce29d49f
rename address to conn_url in a Member obj
...
and introduce new field: api_url
2015-06-02 08:52:34 +02:00
Alexander Kukushkin
bacd05d99e
Determine preferable local address to connect through.
...
If listen contains '*' or 0.0.0.0 - connect via localhost
In all other cases pick the first one.
2015-06-01 17:00:15 +02:00
Alexander Kukushkin
efccd777b2
rename pid_path to postmaster_pid
2015-06-01 16:47:56 +02:00
Alexander Kukushkin
1bcc2b5fa6
Extend postgres?.yml with pg_hba section to give possibility to customize pg_hba.conf
2015-06-01 16:05:35 +02:00
Alexander Kukushkin
adcc7ac256
Try to avoid "double" promotion.
...
Also check presence of trigger_file on master after promotion when
pg_is_in_recovery() = false and if it is there - remove it.
Plus check presence of trigger_file on a new slave (after running
pg_basebackup) and if it is there - also remove it.
2015-06-01 13:48:44 +02:00
Alexander Kukushkin
8c4e547d7a
Handle KeyboardInterrupt in the main loop
2015-06-01 12:02:46 +02:00
Alexander Kukushkin
939254021e
Code cleanup: do not modify os.environ but pass copy of it to subprocess.call
2015-06-01 11:03:06 +02:00
Alexander Kukushkin
f8c3582715
Refactor api:
...
Call is_running only when it's not possible to execute query against
governed postgres.
2015-06-01 10:48:16 +02:00
Alexander Kukushkin
b85262637d
Bugfix: do not create replication slot for master
2015-06-01 09:20:24 +02:00
Alexander Kukushkin
aa486194ae
Wrap time.sleep into wrapper function
...
In case if it was interrupted by SIGCHLD but scheduled awake time is not
reached it will continue sleeping. For all other signals behaviuor is
not changed.
2015-05-28 15:47:51 +02:00
Alexander Kukushkin
cd5e589620
Revert previous commit (restore sigchld_handler)
2015-05-28 13:04:21 +02:00
Alexander Kukushkin
f76f77e964
Ignore SIGCHLD instead of processing it.
...
This way time.sleep wont be interrupted when some external process is
finished and terminated childs would be reaped automatically.
2015-05-28 12:25:49 +02:00
Alexander Kukushkin
01e3beff1c
Merge pull request #2 from CyberDem0n/restapi
...
Restapi
2015-05-27 18:31:42 +02:00
Alexander Kukushkin
bdc4374ae3
connect address for rest api could be different from listen address
2015-05-27 14:51:41 +02:00
Alexander Kukushkin
5a817fad0d
Store in etcd endpoint of restapi
...
In order to have backward compatibility connection url of restapi is
passed as application_name parameter in connection url of database.
Value in etcd will look like:
postgres://user:passwd@host:5432/postgres?application_name=http://host:8080/governor
Sure, this is a hack, but I guess it worth it.
2015-05-27 14:46:35 +02:00
Alexander Kukushkin
77911144c7
Improve test of governor against real config files
2015-05-27 14:44:52 +02:00
Alexander Kukushkin
56a9f52a15
Replace os.system with subprocess.call
2015-05-27 11:16:16 +02:00
Alexander Kukushkin