Motorola 6806800C47B manual Configuration, Time-out Arguments for Checkpoint Service APIs

API Description Time-out Arguments for Checkpoint Service APIs

Creating extra replicas on the system manager node for non-collated checkpoints is an overhead. The advantage of a non-collocated checkpoint is that replica will be created in two places, no matter from how many nodes it is opened.

2.2.2Time-out Arguments for Checkpoint Service APIs

For all synchronous API calls, the application will provide the “timeout” argument. The application will consider invocation of the particular API failed in case it did not complete the call by the specified time. CPSv requires that the value passed in the timeout argument is greater than 100000000 nano seconds (100 milliseconds).

2.2.3Cancellation of Pending Callbacks

According to the SAF-AIS-CKPT-B.01.01 specification, whenever a checkpoint is closed, all the pending callbacks corresponding to this checkpoint should be cancelled. In CPSv, implementation does cancel the pending callbacks related to closed checkpoints. However, the selection object already raised and related to cancelled pending callbacks, will not be cleared or reset. Due to this, saCkptDispatch API may return without invoking callback routine.

2.2.4Maximum Number of Replicas Per Node

CPSv applications can create upto 1,000 replicas per node at a given instance. This includes the replicas created by CPSv for non-collocated checkpoints as per the “replica creation policy.”

In the case of collocated checkpoints, CPSv returns SA_AIS_ERR_NO_RESOURCES if an application attempts to create a new checkpoint and the current number of replicas on the local node is already the maximum that CPSv can support per node.

In the case of non-collocated checkpoints, CPSv returns an SA_AIS_ERR_NO_RESOURCES if the number of checkpoint replicas on the node on which CPSv decides to create a replica is already the maximum that CPSv can support per node. In all other cases, the checkpoint open does not return an error but the replicas will not be created on the backup nodes as decided by the “replica creation policy”.

2.2.5Handling of SA_AIS_ERR_TRY_AGAIN

If the Checkpoint service API returns SA_AIS_ERR_TRY_AGAIN, the application should attempt the API call only after a couple of milliseconds. The suggested wait time is 3 seconds and the number of retries are 12.

Note that the Checkpoint write,overwrite, and read operations may sometime return

SA_AIS_ERR_TRY_AGAIN if called simultaneously. This is to avoid any inconsistencies in the checkpoint database.

2.3Configuration

This section describes how the Checkpoint service is preconfigured regarding shared memory and the maximum write data size.

20 Checkpoint Service Programmer’s Reference (6806800C47B)