NAME | SYNOPSIS | DESCRIPTION | RETURN VALUES | ERRORS | FILES | ATTRIBUTES | SEE ALSO
cc [flags…]-I/usr/cluster/include file -L/usr/cluster/lib -l scha #include <scha.h>scha_err_t scha_control(const char *tag, const char *rgname, const char *rname);
The scha_control() function provides an interface to request the restart or relocation of a resource group or resource that is under the control of the Resource Group Manager (RGM) cluster facility. The command is intended to be used in resource monitors.
The setting of the Failover_mode property of the indicated resource might suppress the requested scha_control action. If Failover_mode is RESTART_ONLY, only SCHA_RESOURCE_RESTART is permitted. Other requests, including SCHA_GIVEOVER, SCHA_CHECK_GIVEOVER, SCHA_RESTART, and SCHA_CHECK_RESTART, return the SCHA_ERR_CHECKS exit code and the requested giveover or restart action is not executed, producing only a syslog message. If the Retry_count and Retry_interval properties are set on the resource, the number of resource restarts is limited to Retry_count attempts within the Retry_interval. If Failover_mode is LOG_ONLY, any scha_control request returns the SCHA_ERR_CHECKS exit code and the requested giveover or restart action is not executed, producing only a syslog message.
The tag argument indicates whether the request is to restart or relocate the resource or group. This argument should be a string value that is defined by one of the following macros, which are defined in <scha_tags.h>:
Perform all the same validity checks that would be done for a SCHA_GIVEOVER of the resource group named by the rgname argument, but do not actually relocate the resource group.
Perform all the same validity checks that would be done for an SCHA_RESTART of the resource group named by the rgname argument, but do not actually restart the resource group.
The SCHA_CHECK_GIVEOVER and SCHA_CHECK_RESTART options are intended to be used by resource monitors that take direct action upon resources, for example, killing and restarting processes, rather than invoking scha_control() to perform a giveover or restart. If the check fails, the monitor should sleep and restart its probes rather than invoke its failover actions. See ERRORS.
The rgname argument is the name of the resource group that is to be restarted or relocated. If the group is not online on the node where the request is made, the request is rejected.
The rname argument is the name of a resource in the resource group. Presumably this is the resource whose monitor is making the scha_control() request. If the named resource is not in the resource group the request is rejected.
The exit code of the command indicates whether the requested action was rejected. If the request is accepted, the function does not return until the resource group or resource has completed going offline and back online. The fault monitor that called scha_control() might be stopped as a result of the resource group's going offline and so might never receive the return status of a successful request.
Requests that the resource group named by the rgname argument be brought offline on the local node, and online again on a different node of the RGM's choosing. Note that, if the resource group is currently online on two or more nodes and there are no additional available nodes on which to bring the resource group online, it can be taken offline on the local node without being brought online elsewhere. The request might be rejected depending on the result of various checks. For example, a node might be rejected as a host because the group was brought offline due to a SCHA_GIVEOVER request on that node within the interval specified by the Pingpong_interval property.
If the cluster administrator configures the RG_affinities properties of one or more resource groups, and you issue a scha_control GIVEOVER request on one resource group, more than one resource group might be relocated as a result. The RG_affinities property is described in rg_properties(5).
The MONITOR_CHECK method is called before the resource group that contains the resource is relocated to a new node as the result of a scha_control(3HA) or scha_control(1HA) request from a fault monitor.
The MONITOR_CHECK method may be called on any node that is a potential new master for the resource group. The MONITOR_CHECK method is intended to assess whether a node is running well enough to run a resource. The MONITOR_CHECK method must be implemented in such a way that it does not conflict with the running of another method concurrently.
MONITOR_CHECK failure vetoes the relocation of the resource group to the node where the callback was invoked.
Requests that failure of the currently executing Prenet_start or Start method should not cause a failover of the resource group, despite the setting of the Failover_mode property.
In other words, this value overrides the recovery action that is normally taken for a resource for which the Failover_Mode property is set to SOFT or HARD when that resource fails to start. Normally, the resource group fails over to a different node. Instead, the resource behaves as if Failover_Mode is set to NONE. The resource enters the START_FAILED state, and the resource group ends up in the ONLINE_FAULTED state, if no other errors occur.
This value is meaningful only when it is called from a Start or Prenet_start method that subsequently exits with a nonzero status or times out. This value is valid only for the current invocation of the Start or Prenet_start method. scha_control() should be called with this value in a situation in which the Start method has determined that the resource cannot start successfully on another node. If this value is called by any other method, the error SCHA_ERR_INVAL is returned. This value prevents the “ping pong” failover of the resource group that would otherwise occur.
Request that the resource restart counter for the resource named by the rname argument be incremented on the local node, without actually restarting the resource.
A resource monitor that restarts a resource directly without calling scha_control() with the RESOURCE_RESTART option (for example, using pmfadm(1M)) can use this option to notify the RGM that the resource has been restarted. This fact is reflected in subsequent scha_resource_get NUM_RESOURCE_RESTARTS queries.
If the resource's type fails to declare the Retry_interval standard property, the RESOURCE_IS_RESTARTED option of scha_control() is not permitted and scha_control() returns error code 13 (SCHA_ERR_RT).
Request that the resource named by the rname argument be brought offline and online again on the local node, without stopping any other resources in the resource group. The resource is stopped and restarted by applying the following sequence of methods to it on the local node:
MONITOR_STOP STOP START MONITOR_START
If the resource's type does not declare a MONITOR_STOP and MONITOR_START method, only the STOP and START methods are invoked to perform the restart.The resource's type must declare a START and STOP method. If the resource's type does not declare both a START and STOP method, scha_control() fails with error code 13 (SCHA_ERR_RT).
If a method invocation fails while restarting the resource, the RGM might either set an error state, relocate the resource group, or reboot the node, depending on the setting of the Failover_mode property of the resource. For additional information, see the Failover_mode property in r_properties(5).
A resource monitor using this option to restart a resource can use the NUM_RESOURCE_RESTARTS query of scha_resource_get() to keep count of recent restart attempts.
The RESOURCE_RESTART function should be used with care by resource types that have PRENET_START or POSTNET_STOP methods. Only the MONITOR_STOP, STOP, START, and MONITOR_START methods are applied to the resource. Network address resources on which this resource implicitly depends is not restarted and remains online.
Request that the resource group named by the rgname argument be brought offline, then online again, without forcing relocation to a different node. The request may ultimately result in relocating the resource group if a resource in the group fails to restart. A resource monitor using this option to restart a resource group can use the NUM_RG_RESTARTS query of scha_resource_get() to keep count of recent restart attempts.
The scha_control() function returns the following values:
The function succeeded.
The function failed.
The function succeeded
The request was rejected. The checks on relocation failed
See scha_calls(3HA) for a description of other error codes.
Normally, a fault monitor that receives an error code from scha_control() should sleep for awhile and then restart its probes, since some error conditions, for example, failover of a global device service causing disk resources to become temporarily unavailable, resolve themselves after awhile. Once the error condition has resolved, the resource itself might become healthy again, or if not, then a subsequent scha_control() request might succeed.
See attributes(5) for descriptions of the following attributes:
ATTRIBUTE TYPE |
ATTRIBUTE VALUE |
---|---|
Availability |
SUNWscdev |
Interface Stability |
Evolving |
rt_callbacks(1HA), scha_control(1HA), scha_calls(3HA), scha_resource_get(3HA), scha_strerror(3HA), attributes(5), r_properties(5), rg_properties(5)
NAME | SYNOPSIS | DESCRIPTION | RETURN VALUES | ERRORS | FILES | ATTRIBUTES | SEE ALSO