How Much Memory Does Your Server Need?
-
Just thought that I would share. This is a real world, production jump box / SSH proxy. It has tuned running which is one of the largest consumers of memory on the system. It also uses filebeat to ship logs to ELK, which is the largest memory consumer on the system. And even with that, while in use, the system uses very, very little memory.
$ free -m total used free shared buff/cache available Mem: 490 62 28162MB of RAM used while running in production. I've seen it idle as low as 39MB!!
-
I was going to start a new thread next week about how to monitor and tweak memory for Linux Servers (and possibly Windows)
-
Simple answer is... shut down everything that you don't use
For example, our Jump box does not, ever, need to send out emails. So the email daemon is disabled. That shaved off 4MB of RAM not having that loaded into memory. We kept it installed, it takes a whole five seconds to turn it on should it ever be needed. But it is not loaded into memory or using CPU cycles. Makes for a lighter, faster, safer system, just a little ... but every little bit helps. -
@scottalanmiller Yeah being a Linux newbie - I'm not sure what commands to use to see what's running, best way to disable etc.
But this is for another days discussion
as I need to sort my Gold Image and Citrix out first lol -
Some quick answers for others interested, if you are using CentOS 7 you can use the systemctl command to list out the daemons set to run at boot time. Here is where you would see postfix or httpd set to fire up with the system.
A common tool to look for big memory users is top. Just run top and select "M" and it will sort the output by the largest memory consumers rather than the largest CPU users (the default view.) If you have a big monitor (or you reduce your font size) you can see dozens of the biggest memory users which generally makes it pretty easy to pick out problem applications and daemons to address or remove.
If you see things like BASH, SSH and top topping the list, you are already too lean to tune it down
-
If you want to run really lean, a good Linux install can get down to around 30MB for a server that is running idle. If you really want, you can get it lower but you have to start removing things that you would normally want on your server, like remote access and the firewall. So realistically you want get much below 55MB without getting a little weird with what you are removing.
-
A Jump box is nearly the leanest thing that you can run. I've seen this one come in at 39MB with me logged in which generates a small amount of memory usage from both SSHD and Bash. Had I collected the number without being logged in, it would have been closer to 38MB. Filebeat alone uses another 4MB and logstash that filebeat needs uses another one or two. If filebeat/logstash was removed, the system might idle around 32MB. That was with Postfix still in the mix, so theoretically closer to 28MB is possible with a truly lean CentOS 7 install. Turn off sysstat collection, don't install fail2ban and you can trim it more, but you are then removing things that I would not recommend removing in most cases.
-
@scottalanmiller Can you post a htop screenshot?
-
Now that the box has been up and running for a bit after a reboot, it cleared down over 10MB of RAM usage!
$ free -m total used free shared buff/cache available Mem: 490 51 -
Here is top:
top - 08:53:22 up 3:48, 1 user, load average: 0.00, 0.01, 0.04 Tasks: 65 total, 2 running, 63 sleeping, 0 stopped, 0 zombie %Cpu(s): 0.0 us, 0.0 sy, 0.0 ni,100.0 id, 0.0 wa, 0.0 hi, 0.0 si, 0.0 st KiB Mem : 501908 total, 244080 free, 50712 used, 207116 buff/cache KiB Swap: 1048572 total, 1048572 free, 0 used. 424180 avail Mem PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 600 root 20 0 553064 16292 5724 S 0.0 3.2 0:03.11 tuned 602 root 20 0 279580 10132 4604 S 0.0 2.0 0:06.68 filebeat 368 root 20 0 450116 8440 6604 S 0.0 1.7 0:02.89 NetworkManager 385 polkitd 20 0 527448 7828 4524 S 0.0 1.6 0:00.13 polkitd 1653 root 20 0 138708 4884 3672 S 0.0 1.0 0:00.08 sshd 607 root 20 0 244008 3916 2572 S 0.0 0.8 0:02.73 logstash-forwar 1 root 20 0 41328 3784 2376 S 0.0 0.8 0:01.78 systemd 351 root 20 0 287416 3628 2920 S 0.0 0.7 0:00.80 rsyslogd 599 root 20 0 82552 3588 2736 S 0.0 0.7 0:00.17 sshd 226 root 20 0 32720 2804 2492 S 0.0 0.6 0:02.41 systemd-journal 386 root 20 0 53060 2680 2100 S 0.0 0.5 0:00.00 wpa_supplicant 1678 scott 20 0 157552 2136 1524 R 0.3 0.4 0:00.03 top 1656 scott 20 0 115508 2096 1628 S 0.0 0.4 0:00.01 bash 1655 scott 20 0 138708 2092 880 S 0.0 0.4 0:00.00 sshd 355 dbus 20 0 26724 1812 1332 S 0.0 0.4 0:00.44 dbus-daemon 257 root 20 0 43028 1748 1296 S 0.0 0.3 0:00.05 systemd-udevd 364 root 20 0 26396 1732 1416 S 0.0 0.3 0:00.28 systemd-logind 365 avahi 20 0 28108 1688 1392 S 0.0 0.3 0:00.16 avahi-daemon 274 root 16 -4 51184 1624 1248 S 0.0 0.3 0:00.17 auditd 382 root 20 0 126328 1584 964 S 0.0 0.3 0:00.12 crond 402 root 20 0 110032 856 732 S 0.0 0.2 0:00.02 agetty 373 avahi 20 0 27984 232 0 S 0.0 0.0 0:00.00 avahi-daemon -
Here is htop
-
Glances for fun...
-
And of course, sar -r
