Voting disk is one of the crucial part in oracle RAC database system and in the absence of it the cluster won't be able to come up. A voting disk is basically containing heart beat information and it contain 2 types of heartbeat i.e. network heartbeat and disk heartbeat.
Disk Heartbeat: Each and every node must store their heartbeat information into the voting disk through which the cssd will make sure that the node is alive in the cluster.
Network Heartbeat: Each and every node must send their heartbeat information to other nodes across the cluster using private interconnect to inform them that he's alive.
When
CSSD agent decides to evict the cluster
If any of the node in the cluster won't be able to store its
heartbeat information into the voting disk then the cssd will validate whether
the node would be able to send their network heartbeat information to other
nodes across the cluster if it found that the node won't be able to send its
heartbeat information to other nodes too then it evicts the node from the
cluster to prevent the split-brain condition.
What information a heartbeat contains
A heartbeat contains only a timestamp nothing more than that.
A split-brain condition is a condition when the culprit node tries
to form an independent cluster within the same cluster and if somehow that
condition will form that it may lead the database’s data into the inconsistent
state which break the ACID property of the oracle database. So, to prevent the
split brain condition cssd agent evict the culprit node from the cluster.
[oracle@oratest2 ~]$ crsctl get css misscount
CRS-4678: Successful get misscount 30 for Cluster
Synchronization Services.
[oracle@oratest2 ~]$ crsctl get css disktimeout
CRS-4678: Successful get disktimeout 200 for Cluster
Synchronization Services.
[oracle@oratest2 ~]$ crsctl query css votedisk
## STATE File Universal Id File Name Disk group
-- ----- ----------------- --------- ---------
1. ONLINE e991f1aa016f4fb7bf5dba810330c3dc
(/dev/oracleasm/disks/ASMDISK01) [DATA]
2. ONLINE 132427f49bc64f63bffcb6bc089813ac
(/dev/oracleasm/disks/ASMDISK02) [DATA]
3. ONLINE 0ac91e3859064f04bfe11ea1b8c78e60
(/dev/oracleasm/disks/ASMDISK03) [DATA]
Located 3 voting disk(s).
In Oracle RAC, each and every node across the cluster must be
able to access more than half of the total number of voting disk which means
more than 50% of the total voting disk must be accessible and if we want to
tolerate the failure of n voting disk in the cluster then the formula is simple
i.e. 2n+1 where n
denotes the total no. of failure we can tolerate. So, for example, if we can tolerate
the failure of 3 voting disk then the total no. of voting disk should be
2*3+1=7, so the formula of 2n+1 always led us to the odd value.
NOTE: The ideal value is total voting
disk is 3.
Probability formula to calculate ideal no. of voting disk
|
Total Number of
Vote Disk |
Failure Probability
of a voting disk |
Probability
of all voting disk failing simultaneously |
Probability that
at least one copy of the voting disk is available |
Marginal Improvement |
|
1 |
0.05 |
0.05 |
0.95 |
- |
|
2 |
0.05 |
0.0025 |
0.9975 |
0.0475 |
|
3 |
0.05 |
0.000125 |
0.999875 |
0.002375 |
|
4 |
0.05 |
0.000006250 |
0.999993750 |
0.000118750 |
|
5 |
0.05 |
0.000000313 |
0.999999688 |
0.000005937 |
|
6 |
0.05 |
0.000000016 |
0.999999984 |
0.000000297 |
