00ce4a504e
Wait for quorum write before returning from Log.Apply().
...
This commit ensures a commit is written to a quorum before returning
from Log.Apply().
2015-05-13 16:05:26 -06:00
f063ebfad2
Default Raft election timeout to 5 seconds
2015-05-12 13:57:28 -07:00
16c8449ca1
Add additional logging for RequestVote.
2015-05-11 15:41:05 -06:00
b8b396807e
Set leader ID on log open if only 1 node
...
Without this change a restart of a single-node cluster would result in
no Raft leader.
2015-05-11 11:23:43 -07:00
3205d8e182
removing unused code
2015-05-04 09:53:03 -06:00
c539072519
Return whether a snap should occur as bool
...
This avoids a race condition when accessing snapshotIndex.
2015-04-30 15:14:25 -07:00
77fea49cca
Use a locking-buffer to prevent races during tests
2015-04-30 14:44:20 -07:00
80a6f2eccd
Do not require snapshot on Log.WriteEntriesTo().
...
This commit fixes a bug where snapshots were always sent when connecting
to the leader -- even when not required.
2015-04-29 14:41:11 -06:00
43827a4ca3
Merge pull request #2423 from influxdb/reenable_elect_test
...
Re-enable leader election test
2015-04-24 15:53:00 -07:00
5894f6580c
Merge pull request #2415 from xiaost/fix-leaderid-error
...
raft: fix leaderID error when state changed
2015-04-24 15:50:33 -07:00
b6fd3605f9
Re-enable leader election test
...
This may have been fixed by PR #2418 .
2015-04-24 15:24:00 -07:00
da5935e885
Fix node getting stuck in candidate state
...
During an election, a node can sometimes get stuck in candidate state
causing it to never read from the new leader. This would prevent
it from incrementing it's index and staying consisistent w/ the
leader.
2015-04-24 16:00:54 -06:00
a9bdfb3b86
Log node state and id in raft traces
...
When running multiple nodes locally w/ raft tracing, it's very difficult
to determine which node is logging. This adds the nodes state(leader,follower,candiate) and
id to all the log messages so we can tarce the nodes states more easily.
2015-04-24 13:25:27 -06:00
a4017fa4fd
Replace heartbeat log threshold time w/ constant
2015-04-24 10:33:34 -06:00
49b80a6d19
Track and log the number of times a heartbeat has failed
2015-04-24 09:44:44 -06:00
2e47a8de11
Log when heartbeat succeeds again
2015-04-24 09:32:15 -06:00
174afb5eb9
Log heartbeat error once every 15s
...
Prevents logs from getting flooded with a message every second when
a broker node is down.
Fixes #1896
2015-04-24 08:43:42 -06:00
e521ac9b73
raft: fix leaderID error when state changed
2015-04-24 22:23:45 +08:00
12c02923c1
Issue 2303
2015-04-15 18:24:33 -07:00
b54f81fcac
Fix stream writer flushing to be thread safe.
...
This change fixes the raft streams so that Flush() is not called
asynchronously while the snapshot is being written.
2015-04-15 12:07:03 -06:00
090f8c7120
fix race
2015-04-15 11:57:03 -06:00
6e124f3ab6
Start cluster before broker.
...
This commit fixes an issue where the second node joins but the first node
cannot commit because it doesn't have the HTTP endpoint running yet. This
is a side effect of streaming raft since we don't synchronize the quorum
set with the heartbeats. We should cache the config per term in the future.
2015-04-15 11:18:44 -06:00
e3e98d67b1
Merge branch 'master' of https://github.com/influxdb/influxdb into raft
...
Conflicts:
CHANGELOG.md
2015-04-13 13:55:30 -06:00
4c4c03852c
Merge branch 'master' of https://github.com/influxdb/influxdb into raft
2015-04-13 13:54:06 -06:00
5750b35b0f
fix chmod is unsupported on the windows
2015-04-13 22:56:27 +08:00
df605fce74
Add log entry cache to raft log.
2015-04-12 16:34:37 -06:00
44b4992b75
Add handling for snapshotting pseudo-state.
2015-04-12 14:24:30 -06:00
908fe7db4a
Wrap raft.Log mutex.
2015-04-12 12:57:34 -06:00
cfea766039
Merge pull request #2213 from cannium/seed
...
Give a seed to rand to make it really random
2015-04-10 16:06:24 -07:00
eaf4bfca0a
Fix term signal.
...
This commit changes raft so that term changes are made immediately and
term change signals are made afterward. Previously, election timeouts
were invalidated by incoming term changes which caused an election loop.
Stale term was also fixed and http/pprof was added too.
2015-04-10 13:52:20 -06:00
c757039f70
Update term change immediately, only send signal.
2015-04-09 17:07:12 -06:00
3f0adcf105
Always notify term change on RequestVote.
...
Previously there was a bug that ignored new terms from incoming vote
requests when the sender's log was out of date.
2015-04-09 13:33:04 -06:00
566f6558b4
Persist term to disk when it changes
2015-04-09 11:12:09 -07:00
7815d47bc3
Merge pull request #2217 from influxdb/cluster-fix
...
Cluster fix
2015-04-09 11:52:52 -06:00
37829ea4c3
Only update term if higher in candidate loop.
...
This fixes a bug where votes were overwritten because terms were
being reset by incoming terms that were the same.
2015-04-09 10:55:39 -06:00
722aa0866f
Give a seed to rand to make it really random
2015-04-09 22:19:55 +08:00
bcfc259fe8
Wait for first applied index on Log.Join().
...
Previously the join method returned immediately after connecting.
Now the join method will wait until the snapshot and/or log entry
is replicated as well before returning.
2015-04-08 13:34:21 -06:00
3d71d45cf6
Return term from vote, add term logging.
2015-04-04 10:59:50 -06:00
6a7cb61f6d
fix unit test is failed with connection refused on the win32
2015-04-03 13:48:52 +08:00
bc444821b9
Add raft term logging.
2015-04-01 15:42:16 -06:00
c326a9c9c9
Fix race condition in Raft unit test
2015-03-30 11:53:27 -07:00
947c776dc8
Fix comments.
2015-03-30 08:13:13 -06:00
189aaa0d82
Remove debug statements.
2015-03-29 15:36:55 -06:00
27f4a3efec
Fix raft transition and reader edge cases.
2015-03-29 15:04:11 -06:00
3d03602fa7
Raft documentation, read from index fix.
2015-03-29 09:08:44 -06:00
057309fc8e
Simplify raft snapshotting, entry apply.
2015-03-26 20:32:39 -06:00
c461c7e7b4
Remove logging to specific file
2015-03-25 17:41:39 -07:00
11c808f55f
Add restore and bootstrap.
...
This commit adds the "influxd restore" command to the CLI. This allows
a snapshot that has been produced by "influxd backup" to be restored
to a config location and the broker and raft directories will be
bootstrapped based on the state of the snapshot.
2015-03-22 15:31:49 -06:00
75450466a6
Fix leader id initialization.
2015-03-16 11:27:57 -06:00
96748cb217
Update file permissions.
2015-03-14 13:49:25 -06:00
Ben Johnson
Philip O'Toole
Cory LaNou
Philip O'Toole
Jason Wilder
xiaost
runner.mei
Can ZHANG