Commits · 7966828a43fc295a7bf63965234962af82a4ba89 · nexedi / neoppod

06 Jul, 2009 21 commits

Delete nodes from the node manager if they are in DOWN state, this is the only · 7966828a

Grégory Wisniewski authored Jul 06, 2009

way to clean old nodes on non-master nodes.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@801 71dcc9de-d417-0410-9af5-da40c76e7ee4

7966828a

Second step for manual startup implementation. Leave the recovering state when · 09eb146e

Grégory Wisniewski authored Jul 06, 2009

the setClusterState packet with is received from the admin node. If no partition
table/IDs were gathered before, wait for at least one storage node and use the
first to create the initial partition table.
A timeout is left to act as it was done before, to preserve automatic startup in
tests, will be removed latter.

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@800 71dcc9de-d417-0410-9af5-da40c76e7ee4

09eb146e

Make handleAnswerPrimaryMaster, handleRequestNodeIdentification and... · 57265e7f

Vincent Pelletier authored Jul 06, 2009

Make handleAnswerPrimaryMaster, handleRequestNodeIdentification and handleAskPrimaryMaster resilient to connection loss. As they are involved in question-response exchanges during election on connections which will get dropped as soon as a master arises, they must check if the connection is still opened before trying to send.
Note: the connection might get closed anyway during the actual send, but this is handled separately already. The problem we fix here emerged because of more tedious checkings in Connection's ask & answer methods.

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@799 71dcc9de-d417-0410-9af5-da40c76e7ee4

57265e7f

When an empty offset list is supplied, answer the whole partition table, to be · a80f94df

Grégory Wisniewski authored Jul 06, 2009

consistent with client <> master communication.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@798 71dcc9de-d417-0410-9af5-da40c76e7ee4

a80f94df

Don't let polling thread die from any exception. Instead, emit a log with... · 3bf17279

Vincent Pelletier authored Jul 06, 2009

Don't let polling thread die from any exception. Instead, emit a log with exception traceback and retry polling.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@797 71dcc9de-d417-0410-9af5-da40c76e7ee4

3bf17279

Add timestamp and log level in log line format string. · 15a469dc
Vincent Pelletier authored Jul 06, 2009
```
git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@796 71dcc9de-d417-0410-9af5-da40c76e7ee4
```
15a469dc

Minors fixes. · fb74cbeb

Grégory Wisniewski authored Jul 06, 2009


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@795 71dcc9de-d417-0410-9af5-da40c76e7ee4

fb74cbeb

First step for manual cluster startup implementation : · a81824a0

Grégory Wisniewski authored Jul 06, 2009

The recovery stage is used only one time, when switching to primary state.
Each time the cluster lost the operational status (a cell has no up to date
node), the verification stage restart.

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@794 71dcc9de-d417-0410-9af5-da40c76e7ee4

a81824a0

Ignore incrementals partition changes during initialization stage, because the · db31b672

Grégory Wisniewski authored Jul 06, 2009

storage is gathering the full table. 


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@793 71dcc9de-d417-0410-9af5-da40c76e7ee4

db31b672

Reset those flags before ask again the informations. · 9a930da9
Grégory Wisniewski authored Jul 06, 2009
```
git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@792 71dcc9de-d417-0410-9af5-da40c76e7ee4
```
9a930da9

When a storage received a node information, it change its server address if it's · 0c84ef82

Grégory Wisniewski authored Jul 06, 2009

already known by UUID. This avoid duplicate entries in the node manager.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@791 71dcc9de-d417-0410-9af5-da40c76e7ee4

0c84ef82

When a node change its address, don't drop it from the partition table node · 64f2ed01

Grégory Wisniewski authored Jul 06, 2009

broadcast it as down state. Just broadcast its information and running state.
Master don't send the partition table to storage node, they are it from the
initialization storage stage.

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@790 71dcc9de-d417-0410-9af5-da40c76e7ee4

64f2ed01

Fix handler class names. · 47046871

Grégory Wisniewski authored Jul 06, 2009


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@789 71dcc9de-d417-0410-9af5-da40c76e7ee4

47046871

Move each handler in its own file. · 33b2bcd9

Grégory Wisniewski authored Jul 06, 2009

Move common handlers in handler.py module.
Add InitializationHandler, applied on the master connection while retreiving
the partition table and node list.
Replace a call to (inexistant) method handleUnexpectedPacket with a raise of
UnexpectedPacketError exception.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@788 71dcc9de-d417-0410-9af5-da40c76e7ee4

33b2bcd9

Fix infinite loop in test_connectToPrimaryMasterNode. · 75c330b2

Vincent Pelletier authored Jul 06, 2009

- test more steps
- fix erroneous assertion
- use a local variable to detect that all handler were executed, rather than a class variable


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@787 71dcc9de-d417-0410-9af5-da40c76e7ee4

75c330b2

Disable by default MySQL queries logging. · 1c569416

Grégory Wisniewski authored Jul 06, 2009


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@786 71dcc9de-d417-0410-9af5-da40c76e7ee4

1c569416

handle 'stopping' cluster state · 64befbf0

Aurel authored Jul 06, 2009


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@785 71dcc9de-d417-0410-9af5-da40c76e7ee4

64befbf0

Fix (some) pyflakes warnings in storage code. · 745c557c

Vincent Pelletier authored Jul 06, 2009


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@784 71dcc9de-d417-0410-9af5-da40c76e7ee4

745c557c

Remove replicator handler from replicator.py, it has moved to · 883bc312

Grégory Wisniewski authored Jul 06, 2009

handlers/replication.py in commit #778.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@783 71dcc9de-d417-0410-9af5-da40c76e7ee4

883bc312

Fix (some) pyflakes warnings in generic code. · b1e46c47

Vincent Pelletier authored Jul 06, 2009


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@782 71dcc9de-d417-0410-9af5-da40c76e7ee4

b1e46c47

typo fix · 0a4a4430

Aurel authored Jul 06, 2009

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@781 71dcc9de-d417-0410-9af5-da40c76e7ee4

0a4a4430

03 Jul, 2009 19 commits

Fix (some) pyflakes warnings in neoctl code. · ad1cbdb7

Vincent Pelletier authored Jul 03, 2009


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@780 71dcc9de-d417-0410-9af5-da40c76e7ee4

ad1cbdb7

Fix (some) pyflakes warnings in master and admin node code. · db8450ca
Vincent Pelletier authored Jul 03, 2009
```
git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@779 71dcc9de-d417-0410-9af5-da40c76e7ee4
```
db8450ca
Add forgotten replication handler (took from replicator.py). · 122585cf
Grégory Wisniewski authored Jul 03, 2009
```
git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@778 71dcc9de-d417-0410-9af5-da40c76e7ee4
```
122585cf

As done for client application, move storage handlers into a dedicated · b75df10f

Grégory Wisniewski authored Jul 03, 2009

directory. Replicator's handler is moved into handlers/replication.py.
Clean some imports.

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@776 71dcc9de-d417-0410-9af5-da40c76e7ee4

b75df10f

Fix imports to avoid test runner crash. · bfc78394

Grégory Wisniewski authored Jul 03, 2009


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@775 71dcc9de-d417-0410-9af5-da40c76e7ee4

bfc78394

Remove all sleeps in testZODB, they are not required any more. · 8c554551
Vincent Pelletier authored Jul 03, 2009
```
git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@774 71dcc9de-d417-0410-9af5-da40c76e7ee4
```
8c554551

Big rework on client handlers: · 4c61df28

Vincent Pelletier authored Jul 03, 2009

- Split completely expected packet handlers from asynchronous event handlers
- Remove NEOStorageConnectionFailure exception and replace it by a more general "ConnectionClosed" exception, purely internal to client app.
- Modify waitMessage to redirect to a handler based on remote peer type
- Fix multiple code paths taken when master doesn't accept connections (or closes them) while trying to conenct to it.
Also:
- remove unused imports found by pyflakes


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@773 71dcc9de-d417-0410-9af5-da40c76e7ee4

4c61df28

Remove more unused imports found by pyflakes. · a0f2936c

Vincent Pelletier authored Jul 03, 2009


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@772 71dcc9de-d417-0410-9af5-da40c76e7ee4

a0f2936c

Remove unused imports found by pyflakes. · aec0cb00

Vincent Pelletier authored Jul 03, 2009


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@771 71dcc9de-d417-0410-9af5-da40c76e7ee4

aec0cb00

One more undefined STORAGE_NODE_TYPE usage (forgotten in Gregosy's recent commit). · e1905079
Vincent Pelletier authored Jul 03, 2009
```
git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@770 71dcc9de-d417-0410-9af5-da40c76e7ee4
```
e1905079

Make MTClientConnection.close expect to be called only when protected by a lock. · 9b0b3fb3

Vincent Pelletier authored Jul 03, 2009

Add lock/release around each "close" call in client which wasn't already locked.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@769 71dcc9de-d417-0410-9af5-da40c76e7ee4

9b0b3fb3

Fix attribute name and constant access. · c797d830

Grégory Wisniewski authored Jul 03, 2009


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@768 71dcc9de-d417-0410-9af5-da40c76e7ee4

c797d830

Remove a dangling remain of connection message order change. · 6e0075ea
Vincent Pelletier authored Jul 03, 2009
```
git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@767 71dcc9de-d417-0410-9af5-da40c76e7ee4
```
6e0075ea

Fix exception name. · a0c8a7d9

Grégory Wisniewski authored Jul 03, 2009


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@766 71dcc9de-d417-0410-9af5-da40c76e7ee4

a0c8a7d9

Split client handlers into 3 parts: · a4e7ac97

Vincent Pelletier authored Jul 03, 2009

- generic base handler
- client connection handlers
- master connection handlers


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@765 71dcc9de-d417-0410-9af5-da40c76e7ee4

a4e7ac97

Move handler.py into a separate folder. · b6034e79

Vincent Pelletier authored Jul 03, 2009

Update users.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@764 71dcc9de-d417-0410-9af5-da40c76e7ee4

b6034e79

Replace an "from ... import *" by explicit import list. · d0fa630a
Vincent Pelletier authored Jul 03, 2009
```
git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@763 71dcc9de-d417-0410-9af5-da40c76e7ee4
```
d0fa630a

Typo in variable name. · 0f4e1c76

Vincent Pelletier authored Jul 03, 2009


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@762 71dcc9de-d417-0410-9af5-da40c76e7ee4

0f4e1c76

Make notifyDeadStorage callable for any kind of node (and rename it to notifyDeadNode). · 02fea1d5

Vincent Pelletier authored Jul 03, 2009

For now, it's a no-op for any non-storage node.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@761 71dcc9de-d417-0410-9af5-da40c76e7ee4

02fea1d5