A write fail warning


#1

My cluster was configured as all data in memory with HDD for persistence. “WARNING (drv_ssd): (drv_ssd.c::4260) {state} write fail: queue too deep: q 65, max 64” is found in the log when I were running workload. What’s that mean?


#2

This means the disk writes are taking too long (the Aerospike write queue is full and is not being emptied fast enough). The underlying reason is simply that you have reached the write throughput of your disk.


#3

You will find more details on this here: device overload troubleshooting.