I added one new node to a cluster for the purpose of lowering the memory used per node, however, after the node is added, the free namespace memory for the existing nodes did increase, but the free system memory for the existing nodes are staying the same, I am wondering what is going on within the Aerospike? Is Aerospike holding this memory for future data?
This is the system memory info:
total used free shared buff/cache available
Mem: 259999656 209928584 12126304 36248184 37944768 8061072
Swap: 0 0 0
And the namespace memory info:
Namespace| Node| Total|Expirations|Evictions| Stop|~Disk~|~~~~~~~~~~~Memory~~~~~~~~~~|~Primary~
| | Records| | |Writes| HWM%| Used|Used%|HWM%|Stop%|~~Index~~
| | | | | | | | | | | Type
xxx |aaa.bbb.ccc.ddd |488.466 M| 604.144 M| 0.000 |False | 0|177.643 GB| 81| 100| 100|shmem
xxx |aaa.bbb.ccc.ddd |488.737 M| 4.664 G| 0.000 |False | 0|177.173 GB| 81| 100| 100|shmem
xxx |aaa.bbb.ccc.ddd |488.785 M| 4.665 G| 0.000 |False | 0|177.188 GB| 81| 100| 100|shmem
xxx |aaa.bbb.ccc.ddd |488.336 M| 4.666 G| 0.000 |False | 0|177.041 GB| 81| 100| 100|shmem
xxx |aaa.bbb.ccc.ddd |488.539 M| 4.662 G| 0.000 |False | 0|177.078 GB| 81| 100| 100|shmem
xxx | | 2.443 G| 19.261 G| 0.000 | | |886.124 GB| | | |
And this is the namespace configuration:
namespace xxx {
replication-factor 2
memory-size 220G
nsup-period 5m
default-ttl 1825D
high-water-memory-pct 100
storage-engine memory
stop-writes-pct 100
transaction-pending-limit 500
single-scan-threads 8
conflict-resolution-policy last-update-time
}
I also have the metrics from AMS: