New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
<dustin> Thank you for your support, your slide decks have been doubled.
I gave him extra RAM and I lost my job. - Sreejit
I wouldn't mind losing that tiny bit of extra RAM over someones job.
Check
dmesgand/var/log/kern.logtoo, particularly the messages at boot. It should usually say if some amount of RAM is reserved for the system.Yes, it was. Definitely a 1D10T error on my part for not properly investigating. I am used to debug thoroughly, but lately I got accustomed to move fast and break things.
I paid for a 2.5 GB VPS, but the "free -h" command says they gave me only "2.4 Gi" of memory.
Should I complain? ;-)
yes
Not an issue, but you may want to check if kdump, as a service, eats up your RAM.
Ask GPT, and she knows well about that.
GPT is a male:
Sure. You will get 2.6GB if you are lucky.
Here's the available memory on a 1G KVM VPS running AlmaLinux 8, without the kdump space reserved:
[rk@p3 ~]$ free -m
total used free shared buff/cache available
Mem: 964 136 545 7 282 683
Swap: 998 0 998
The reserved space for kdump did take up a chunk of memory. I don't remember exactly how much, but I seem to remember that the total available with kdump was about 840 GB, instead of the current 964. But I hadn't ever used kdump (in at least 10 or so years), and so disabled it on my VPSes. I can re-enable it if I ever need it, which will probably be approximately never. YMMV.
It's a fairly small amount of extra RAM, 124 MB or so, but someone with a 512MB or 768MB VPS might want to take the trouble to configure it. For 1GB and up, hmm, probably not too much difference, as long as you have enough swap (I like Zram).
This should do it, as in the earlier cited article:
sudo systemctl status kdump
sudo systemctl stop kdump
sudo systemctl disable kdump
sudo nano /etc/default/grub
In /etc/default/grub, find the line that starts with
GRUB_CMDLINE_LINUX="crashkernel=auto ... [etc.]
and change it to
GRUB_CMDLINE_LINUX="crashkernel=no ... [etc.]
Then do:
sudo grub2-mkconfig -o /boot/grub2/grub.cfg
sudo systemctl reboot
Sooo, kdump. Is this indicate lack of care or just plain incompetence from provider side? I mean, templates should be optimized!
Complain, get free ram, profit
True confession:
I was teasing. The "question" was really about the difference between metric prefixes (e.g., giga-, mega-, etc.) and binary prefixes (e.g., gibi-, mebi-, etc.). The metric prefixes are slightly smaller values compared with their binary equivalents.
Metric prefixes:
1 GB = 1 gigabyte = 1000 million bytes = 1000000000 bytes.
1 MB = 1 megabyte = 1 million bytes = 1000000 bytes.
With the binary prefixes:
1 GiB = 1 gibibyte = 1024^3 bytes = 1073741824 bytes.
1 MiB = 1 mebibyte = 1024^2 bytes = 1048576 bytes.
I assume that most VPS providers sell in metric GB units = 1000 million bytes. The providers deliver slightly less RAM than if they used binary prefix units, such as GiB.
The Linux "free" command gives results in binary prefixes, which explains the difference in my example. I chose "free -m" (mebibytes) because "free -g", "free --gibi" and "free --giga" all yield the same results for my 2.5 GB VPS: "2".
I hope that clears it all up. Just teasing.
Related Question:
Has the world finally standardized on what "1 billion" means? For a while it varied in different places in the world:
US: thousand - million - billion - trillion
UK: thousand - million - 1000 million - billion - ???
Has that changed? Has the rest of the world migrated to the US usage today? ... or are we still in the "pay attention to the source" mode?
The UK traditionally used the French names:
UK: thousand - million - milliard - billion - billiard - trillion - trilliard
Although the -iard forms were dying out quite a while ago and we generally preferred thousand -ion instead, but you would see the -iard ones from time to time.
Nowadays, most financial use in the UK follows the US for billion, but most normal people say thousand million instead of billion. But then most normal people also don't often need numbers that big.
Nyah. I've done the same a million times.
If only the world had the metric system to sort out these problems.
ha ha ha ha ha , true............
I just ran across an old file that recorded the total available RAM with kdump enabled as 804 MB.
So the actual gain in memory is about 160 MB. Sorry for the confusion.
I know this is necro but on the other hand I don't want to open a new thread on this same topic. I'll let the mods decide if it was the wrong move
I realised recently that my RHEL 9.4 VPS with 2G RAM was showing 1.7G in free instead of the advertised 2G. Now I did change the GRUB_CMDLINE_LINUX="crashkernel=no" thing in /etc/default/grub a year ago after reading this thread, but it was no longer working (did it ever work, I'm not sure).
It turns out the crashkernel argument is defined in the entries in /boot/loader/entries/.
After some research I discovered that there is a tool called grubby which oversees creating these entries.
I typed:
grubby --args=crashkernel=no --update-kernel /boot/vmlinuz-5.14.0-427.31.1.el9_4.x86_64After reboot it worked.
I'm not sure if I will have to do the same every time a new kernel is installed??