For this use case of changing the I/O mode to recover from a suspended GAD pair, operations are stopped due to a failure in a storage system. The P-VOL in the failure site requires time for recovery because the pair status and I/O mode cannot be checked. The S-VOL is suspended with I/O mode Block, so operations cannot resume.
This use case can also be applied when the status of the P-VOL and S-VOL is opposite from what is described here. If this is the case, read P-VOL as S-VOL and S-VOL as P-VOL in the following recovery procedure.
For the P-VOL, delta resync might be already performed in a GAD+UR environment. If this is the case, perform the optional deletion step on the UR pair and the UR delta resync pair, restore the GAD pair, and then perform the reconfiguration step on the UR pair and the UR delta resync pair according to the following procedure.
-
Switch the I/O mode of the S-VOL to Local to resume operations.
Command example when S-VOL is instance 1:
pairsplit -g oraHA -iomd local -IH1
The I/O mode of the S-VOL is forcibly switched from Block to Local, and then you can resume operations. The pair status of the S-VOL changes from PSUE to SSWS. The data might not be the latest one because the data is as of when the GAD pair was suspended.
-
Before starting recovery operation on the failure site, take the host path
from the host to the P-VOL offline.
CAUTION:Ensure that you take the host path to the P-VOL offline. If you do not take it offline, split-brain might occur on the host if the I/O mode of the P-VOL was Local before the failure occurred.
-
After the failure is recovered, if the P-VOL is suspended with I/O mode Local,
forcibly switch the I/O mode of the P-VOL from Local to Block before the host
recognizes the P-VOL.
Command example when the P-VOL is instance 0:
pairsplit -g oraHA -iomd block -IH0
-
Bring the host path online.
-
Resynchronize the pair by specifying the S-VOL (on which operations resumed) to
recover the GAD pair status.
When the pair is resynchronized, initial copy (copying of the entire data, which is the same operation as copying when a pair is created) might be performed, depending on the state when a failure occurs.