TCP and sock_inode_cache slabs are increasing on the NFS client system while the NFS server down

Issue

• While the NFS server down, TCP and sock_inode_cache slabs are increasing on the NFS client system.
• Once NFS server responds, these slabs are freed.

MemTotal:        3820748 kB
MemFree:          166616 kB
Active:           574072 kB
Inactive:         986052 kB
Slab:            1641476 kB   <<<--- 42% of Total Memory
SReclaimable:     692140 kB
SUnreclaim:       949336 kB   <<<--- high
Percpu:             1936 kB

TCP and sock_inode_cache consume 56% of Slab.

$ cat proc/slabinfo |awk '{print $1, $2*$4/1024 " KB"}'|sort -k2nr
TCP 721966 KB
xfs_inode 223227 KB
sock_inode_cache 203902 KB
...