Could you please share the information related to fence level setting for paircreate?
how to identify what could be the fence level & which is best level as per the system requirement?
To understand fence level, you need to think it from the perspective of "what happens when failure occurs?" . I think this should give you good explanation of 'fence level'
Fence Level options for I/O to the P-VOL after suspension:
The P-VOL Fence Level setting determines whether the host is denied access or continues to access the P-VOL when the pair is suspended due to an error.
You specify one of the following Fence Level options during the initial copy and resync operations. You can also change the Fence Level option outside these operations.
• Data – the P-VOL is fenced if an update copy operation fails. This prevents the host from writing to the P-VOL during a failure. This setting should be considered for the most critical volumes for disaster recovery.
This setting reduces the amount of time required to analyze the currency of the S-VOL during disaster recovery efforts. This setting is also designed for applications that can continue to operate with another device pair.
• Status – the P-VOL is fenced only if the primary system is not able to change S-VOL status to suspended when an update copy operation fails.
If the primary system successfully changes S-VOL pair status to suspended, subsequent write I/O operations to the P-VOL will be accepted, and the system will keep track of updates to the P-VOL. This allows the pair to be resynchronized quickly. This setting also reduces the amount of time required to analyze S-VOL currency during disaster recovery.
• Never – the P-VOL is never fenced. This setting should be used when I/O performance out-weighs data recovery. "Never" ensures that the P-VOL remains available to applications for updates, even if all TrueCopy copy operations have failed. The S-VOL may no longer be in sync with the PVOL, but the primary system keeps track of updates to the P-VOL while the pair is suspended. Host failover capability is essential if this fence level setting is used. For disaster recovery, the currency of the S-VOL is determined by using the sense information transferred via host failover or
by comparing the S-VOL contents with other files confirmed to be current.
→From the options you have available, you would need to know your application to select the best option.
→hope this is helpful.
Here's a out-of-the-manual explanation.
Fence level is always set to Never when you are replicating File systems.
Fence level is set to data for databases to avoid data corruption during a rolling disaster.
A rolling disaster is the time period between the start of a disaster and the end of a disaster.
During this period you could have dropped FC frames or exchanges or a out of order delivery that
violates the database log ahead protocol at the DR site.
Consider that you are a business that trades in high volumes on the stock exchange.
If there is a disaster then you could potentially accept transactions that may not get reflected in the DR site if the Production site goes down after the replication links go down during the rolling disaster period. If after the disaster the Production Data is completely unrecoverable then even if you have accepted 5 minutes worth of transactions that are missing on the DR site it could mean millions of dollars in lost transactions.
This is one scenario where you may want to set the Fence level to data. That is if a transaction cannot be mirrored on the remote site then it is not committed to the production site since the production volumes will be fenced off or turned to read only.
The downside of this is that if you lose the link between the Production Site and the DR site then you have a unwanted production outage.
Sometimes customers separate the database and log consistency groups and set Fence Level to Never for the database and set Fence Level to Data for the log volumes. Then you have to configure the database to use a alternate log volume when the primary log volumes are unavailable. This helps with link outages not causing unnecessary production outages and with the avoiding the ever present threat of data corruption during a rolling disaster.
Retrieving data ...