We have write-q in case of storage-engine = device, so we can monitor the max-write-cache. But if we use pmem as storage engine, we don’t have write-q for that.
So as the documentation suggest we can configure max-write-cache for pmem. how we can monitor it.
pmem storage uses 8M write-block-size, not configurable by user. With client writes, defrag writes, migration writes, replica writes, each using their own 8M current block, just from a write-block management point of view, the max-write-cache may be used internally by the code path but the writes are actually happening on the pmem. Leave max-wirte-cache setting at default, there is no process ram being allocated to max-write-cache, and you should never see device_overload error. (In a future release, max-write-config for pmem may be internally set and removed as a user config option.)