0

I wanted to have an opinion from anybody about my production server swapiness and physical ram sharing into services.

This server configuration has OS 22.04 ubuntu with 128GB RAM. It is a dedicated server not a vps. It hosts a php application and run the following services:

  1. Nginx 1.26.2
  2. Varnish 7.5 run in socket
  3. Redis 7.4.1 run in socket
  4. PHP8.3 FPM
  5. Rabbitmq 3.13.7
  6. ElasticSearch 8.11.4
  7. Percona mysql 8.0.39-30
  8. Hitch 1.8

All above services have of course custom configurations which included the amount of memory should be used. For this server I have configure a swap file of 64GB. Of course I have given no priority to system to use this SWAP

NAME      TYPE      SIZE USED PRIO
/swapfile file       64G 2.8G   -2
/dev/md0  partition   4G   0B   -3

cat /proc/sys/vm/swappiness
10

and

/etc/sysctl.conf
vm.swappiness=10
vm.vfs_cache_pressure=90
vm.overcommit_memory = 1

RAM free -m

               total        used        free      shared  buff/cache   available
Mem:          128731       46517        7285         953       74928       79952
Swap:          69626        2882       66744

You will notice that I have around 2.8GB of swap used. As far as I know OS to use SWAP must have full RAM.

With this bash script:

for file in /proc/*/status; do awk '/VmSwap|Name/{printf $2 " " $3}END{ print ""}' $file; done | sort -k 2 -n -r | less

I see that nginx, php8.3, Elasticsearch, redis and hitch use SWAP

Also I see in RAM that there is around 7GB free. Also I see that buff/cache uses 75GB of RAM. Here come the questions:

  1. How I can prevent OS to use SWAP?
  2. Can someone explain what is buff/cache? Is this reserve RAM. Is this related with Varnish? How I control this section of RAM?
G. G.
  • 155

0 Answers0