Dell PowerMax Operate (DCA Operate): Remote Replication (SRDF)

SRDF questions test whether you know which mode gives which recovery point, and the exact order of operations in a failover or failback.

Topology and terms

• The R1 side is the source, the R2 side the target. Pairs are organized into RDF groups, and the same group number must exist on both arrays.
• RDF groups carry the RA (remote adapter) ports; a group is defined on each array before pairs are created.
• Consistency groups preserve dependent-write ordering, which is what makes a recovered copy usable for databases.

Modes

• Synchronous: the host write is acknowledged only after the R2 write completes. Zero data loss within the pair, but latency depends on the link.
• Semi-synchronous: acknowledgement after the R2 write is accepted into cache; less latency-sensitive than synchronous.
• SRDF/A (asynchronous): writes are shipped in delta sets with cycle switching. RPO is up to one cycle plus transmit lag, not zero.
• Adaptive copy (disk / write pending): a data-mover mode used to catch up a link; it does not guarantee an RPO, so it is not a production steady state.
• SRDF/Metro: both arrays present the same devices read/write to the host, using bias and a witness to keep the pair consistent.

Core commands

symrdf addgrp to define the group, symrdf createpair -sid 123 -rdfg 5 -type rdf1 -mode sync -sg APP_SG -remote_sid 456 -remote_rdfg 5, then symrdf establish, symrdf query, symrdf suspend, symrdf resume, symrdf swap, symrdf failover, symrdf failback, symrdf set mode.

Worked example

DR test on a synchronous group: (1) symrdf -sid 123 -rdfg 5 query confirms pairs are Synchronized; (2) symrdf suspend moves the pairs to Suspended; (3) symrdf failover makes the R2 devices read/write and the test host mounts them; (4) after testing, symrdf failback and symrdf establish resynchronize. Issuing failover while pairs are still Synchronized fails, because the target must be suspended before it can be made write-enabled. If writes were made on both sides during the test, you need a full establish, not an incremental one.

Traps

• Forgetting to create the matching RDF group on the remote array.
• Quoting SRDF/A RPO as zero.
• Leaving adaptive copy mode on a production group.
• Incremental resume after divergent writes on both sides, which corrupts consistency.
• Swapping R1/R2 in commands with the wrong -type direction.