Database Reference
In-Depth Information
ora.evmd
1 ONLINE ONLINE ssky3l11p1 STABLE
ora.gipcd
1 ONLINE ONLINE ssky3l11p1 STABLE
ora.gpnpd
1 ONLINE ONLINE ssky3l11p1 STABLE
ora.mdnsd
1 ONLINE ONLINE ssky3l11p1 STABLE
ora.storage
1 ONLINE ONLINE ssky3l11p1 STABLE
--------------------------------------------------------------------------------
The Cluster Synchronization Service (CSS)
CSS is a subcomponent of Oracle Clusterware. It maintains membership in the cluster through a special file called
a voting disk (also referred to as a quorum disk), which is also on a shared-storage subsystem visible to all nodes
participating in the cluster.
[oracle@ssky3l11p1 ~]$ $GRID_HOME/bin/crsctl query css votedisk
## STATE File Universal Id File Name Disk group
-- ----- ----------------- --------- ---------
1. ONLINE 8ee93e68dfdb4fdfbf20036e0e42af51 (ORCL:AVOL1) [GRID_DATA]
2. ONLINE 82506bd06acf4f3bbfcb7dc6a89e256e (ORCL:AVOL2) [GRID_DATA]
3. ONLINE b43b8039540f4f03bfee5a99b06bd918 (ORCL:AVOL3) [GRID_DATA]
Located 3 voting disk(s).
Oracle Clusterware Stack
At the cluster level, the main processes of Oracle Clusterware provide a standard cluster interface on all platforms
and perform high availability operations on each node in the cluster. Figure
16-1
illustrates the various processes that
compose the Oracle GI stack.
Initiated by the CSS process after nodes start up, the Oracle Cluster Synchronization Service Daemon, or CSSD, is
a process that performs basic synchronization services between the various resources in the cluster. With the help of
the voting disk, it arbitrates ownership of the cluster among cluster nodes in the event of a complete private network
failure. CSSD is a critical daemon process and a failure of this process causes the node (server) to reboot.
cssdagent
(Figure
16-1
, section 1) is a new agent process introduced in Oracle Clusterware 11g Release 2 to
start the CSSD process and tracks its activities by logging into the agent directory within the
$GRID_HOME/log/
<node name>/agent/
directories.
cssdmonitor
(Figure
16-1
, section 1) is responsible for monitoring the health of the CSSD and replaces the
OPROCD process that existed in the earlier versions of RAC.
Both
cssdagent
and
cssdmonitor
track the state information from the CSSD. These CSS services are performed
by the node membership (NM) service and the group membership (GM) services.
The NM checks the heartbeat across the various nodes in the cluster every second across the private network
(network heartbeat, or NHB). It also alternates to check the disk heartbeat (DHB) by performing a read/write
operation every second to the voting file. If the heartbeat/node members do not respond within 30 seconds, as
defined by the miscount parameter, the node (among the surviving nodes) that was started first (a.k.a., the master)
will start evicting the other node(s) in the cluster.
Search WWH ::
Custom Search