If an I/O issued to ONTAP requires data from the capacity tier and ONTAP cannot reach the capacity tier to retrieve blocks, then the I/O eventually times out. The effect of this timeout depends on the protocol used. In an NFS environment, ONTAP responds with either an EJUKEBOX or EDELAY response, depending on the protocol. Some older operating systems might interpret this as an error, but current operating systems and current patch levels of the Oracle Direct NFS client treat this as a retriable error and continue waiting for the I/O to complete.

A shorter timeout applies to SAN environments. If a block in the object store environment is required and remains unreachable for two minutes, a read error is returned to the host. The ONTAP volume and LUNs remain online, but the host OS might flag the file system as being in an error state.

Object storage connectivity problems snapshot-only policy is less of a concern because only backup data is tiered. Communication problems would slow data recovery but would not otherwise affect data being actively used. The auto and all policies allow tiering of cold data from the active LUN, which means that an error during object store data retrieval could affect database availability. A SAN deployment with these policies should only be used with enterprise-class object storage and network connections designed for high availability. NetApp StorageGRID is the superior option.