27289 Commits

Author SHA1 Message Date
antirez
16bc1ae5f4 Sentinel: failover restart time is now multiple of failover timeout.
Also defaulf failover timeout changed to 3 minutes as the failover is a
fairly fast procedure most of the times, unless there are a very big
number of slaves and the user picked to configure them sequentially (in
that case the user should change the failover timeout accordingly).
2013-11-18 11:30:08 +01:00
antirez
83316f515c Sentinel: failover restart time is now multiple of failover timeout.
Also defaulf failover timeout changed to 3 minutes as the failover is a
fairly fast procedure most of the times, unless there are a very big
number of slaves and the user picked to configure them sequentially (in
that case the user should change the failover timeout accordingly).
2013-11-18 11:30:08 +01:00
antirez
7b7763ff3e Sentinel: state machine and timeouts simplified. 2013-11-18 11:12:58 +01:00
antirez
3a56013acb Sentinel: state machine and timeouts simplified. 2013-11-18 11:12:58 +01:00
antirez
00cad98228 Sentinel: election timeout define. 2013-11-18 10:08:06 +01:00
antirez
4be53b1c5d Sentinel: election timeout define. 2013-11-18 10:08:06 +01:00
antirez
297df0e910 Sentinel: fix address of master in Hello messages.
Once we switched configuration during a failover, we should advertise
the new address.

This was a serious race condition as the Sentinel performing the
failover for a moment advertised the old address with the new
configuration epoch: once trasmitted to the other Sentinels the broken
configuration would remain there forever, until the next failover
(because a greater configuration epoch is required to overwrite an older
one).
2013-11-14 10:25:55 +01:00
antirez
69d826a354 Sentinel: fix address of master in Hello messages.
Once we switched configuration during a failover, we should advertise
the new address.

This was a serious race condition as the Sentinel performing the
failover for a moment advertised the old address with the new
configuration epoch: once trasmitted to the other Sentinels the broken
configuration would remain there forever, until the next failover
(because a greater configuration epoch is required to overwrite an older
one).
2013-11-14 10:25:55 +01:00
antirez
a031dc6efe Sentinel: master address selection in get-master-address refactored. 2013-11-14 10:23:54 +01:00
antirez
e4c65e72c6 Sentinel: master address selection in get-master-address refactored. 2013-11-14 10:23:54 +01:00
antirez
d1864486fc Sentinel: fix conditional to only affect slaves with wrong master. 2013-11-14 10:23:05 +01:00
antirez
c0d7229364 Sentinel: fix conditional to only affect slaves with wrong master. 2013-11-14 10:23:05 +01:00
antirez
f4216da4f1 Sentinel: simplify and refactor slave reconfig code. 2013-11-14 00:36:43 +01:00
antirez
dfbd9c5aeb Sentinel: simplify and refactor slave reconfig code. 2013-11-14 00:36:43 +01:00
antirez
5ceb9dc93e Sentinel: reconfigure slaves to right master. 2013-11-14 00:29:38 +01:00
antirez
64ad6648a8 Sentinel: reconfigure slaves to right master. 2013-11-14 00:29:38 +01:00
antirez
e7fb6c5697 Sentinel: remember last time slave changed master. 2013-11-14 00:20:15 +01:00
antirez
3e27d678da Sentinel: remember last time slave changed master. 2013-11-14 00:20:15 +01:00
antirez
4608f5da94 Sentinel: redirect-to-master is not ok with new algorithm.
Now Sentinel believe the current configuration is always the winner and
should be applied by Sentinels instead of trying to adapt our view of
the cluster based on what we observe.

So the only way to modify what a Sentinel believe to be the truth is to
win an election and advertise the new configuration via Pub / Sub with a
greater configuration epoch.
2013-11-13 17:03:48 +01:00
antirez
8297745fa6 Sentinel: redirect-to-master is not ok with new algorithm.
Now Sentinel believe the current configuration is always the winner and
should be applied by Sentinels instead of trying to adapt our view of
the cluster based on what we observe.

So the only way to modify what a Sentinel believe to be the truth is to
win an election and advertise the new configuration via Pub / Sub with a
greater configuration epoch.
2013-11-13 17:03:48 +01:00
antirez
f81306af33 Sentinel: safer slave reconfig, master reported role should match. 2013-11-13 17:02:09 +01:00
antirez
76a88f56e5 Sentinel: safer slave reconfig, master reported role should match. 2013-11-13 17:02:09 +01:00
antirez
406fee011d Sentinel: role reporting fixed and added in SENTINEL output. 2013-11-13 16:39:57 +01:00
antirez
ddaad9fe2d Sentinel: role reporting fixed and added in SENTINEL output. 2013-11-13 16:39:57 +01:00
antirez
cc721419ac Sentinel: being a master and reporting as slave is considered SDOWN. 2013-11-13 16:28:56 +01:00
antirez
a0afa66f4b Sentinel: being a master and reporting as slave is considered SDOWN. 2013-11-13 16:28:56 +01:00
antirez
b27dd8b9d3 Sentinel: make sure role_reported is always updated. 2013-11-13 16:21:58 +01:00
antirez
17718fdcba Sentinel: make sure role_reported is always updated. 2013-11-13 16:21:58 +01:00
antirez
72629ad358 Sentinel: track role change time. Wait before reconfigurations. 2013-11-13 16:18:23 +01:00
antirez
46a053d34b Sentinel: track role change time. Wait before reconfigurations. 2013-11-13 16:18:23 +01:00
antirez
e1d06fdeb9 Sentinel: fix no-down check in master->slave conversion code. 2013-11-13 13:43:59 +01:00
antirez
9e40c46f5e Sentinel: fix no-down check in master->slave conversion code. 2013-11-13 13:43:59 +01:00
antirez
47598f4a88 Sentinel: readd slaves back after a master reset. 2013-11-13 13:01:11 +01:00
antirez
ae35b7e240 Sentinel: readd slaves back after a master reset. 2013-11-13 13:01:11 +01:00
antirez
f5b5a8eb9f Sentinel: sentinelResetMaster() new flag to avoid removing set of sentinels.
This commit also removes some dead code and cleanup generic flags.
2013-11-13 10:30:45 +01:00
antirez
6bd4f6bffe Sentinel: sentinelResetMaster() new flag to avoid removing set of sentinels.
This commit also removes some dead code and cleanup generic flags.
2013-11-13 10:30:45 +01:00
antirez
9896fd51e0 Sentinel: receive Pub/Sub messages from slaves. 2013-11-12 23:07:33 +01:00
antirez
1569af1f23 Sentinel: receive Pub/Sub messages from slaves. 2013-11-12 23:07:33 +01:00
antirez
37203602f9 Sentinel: change event name when converting master to slave. 2013-11-12 23:00:17 +01:00
antirez
dfa5f8b777 Sentinel: change event name when converting master to slave. 2013-11-12 23:00:17 +01:00
antirez
4d365131b0 Sentinel: added config-epoch to SENTINEL masters output. 2013-11-12 17:22:04 +01:00
antirez
24158d1488 Sentinel: added config-epoch to SENTINEL masters output. 2013-11-12 17:22:04 +01:00
antirez
1e74623446 Sentinel: new failover algo, desync slaves and update config epoch. 2013-11-12 17:07:31 +01:00
antirez
d2bc6dc39a Sentinel: new failover algo, desync slaves and update config epoch. 2013-11-12 17:07:31 +01:00
antirez
61794132b2 Sentinel: when starting failover seek for votes ASAP. 2013-11-12 16:38:02 +01:00
antirez
4a128b949d Sentinel: when starting failover seek for votes ASAP. 2013-11-12 16:38:02 +01:00
antirez
d63c15923f Sentinel: +new-epoch events. 2013-11-12 13:35:25 +01:00
antirez
e6b9d5e97e Sentinel: +new-epoch events. 2013-11-12 13:35:25 +01:00
antirez
ed01b141e8 Sentinel: wait some time between failover attempts. 2013-11-12 13:30:31 +01:00
antirez
54c447be52 Sentinel: wait some time between failover attempts. 2013-11-12 13:30:31 +01:00