Known Issues

DTS2022121317561

In a capacity expansion scenario where five nodes are added to a five-node cluster, inject a fault to the nodes to be added to construct a capacity expansion failure. Perform capacity expansion again without formatting the nodes to be added. In this case, the capacity expansion scenario changes (three nodes to be added to a five-node cluster), and incorrect I/Os occasionally occur.

Minor

If a node to be added is faulty, format the cache drive of the faulty node before the second capacity expansion to ensure that the number of nodes in the first capacity expansion is the same as that in the second capacity expansion. This issue is clearly stated in the feature guide.

This issue will be resolved in version RC2.B100.

1. If a fault is injected during capacity expansion and the capacity expansion is continued after the faulty node is recovered without being formatted, there is a possibility (about 15%) that incorrect I/Os occur. Error I/Os are not written, which does not affect data correctness on the server. I/Os can be properly executed after being resent.
2. Currently, no Global Cache cluster exists on the live network, and capacity expansion is not required. Therefore, this problem will not be triggered on the user plane.
3. This problem is caused by proactive fault injection during capacity expansion. In actual capacity expansion practice, the probability of faults is low.
4. Impact on services: Services can be properly executed after initialization. The original service data is not affected, and data consistency is maintained.