Join Log in
Score:0
rc1 avatar Server

Clarification on understanding Linux disk usage with df and how tmpfs should be assessed

de flag
rc1

I have searched serverfault, SO and other sites but could not find a clear answer. I have done some reading on the basics of Linux storage and filesystems, but I'm still unclear about how to solve my problem.

My aim is to do a simple assessment of disk space and usage across servers in our environment. We will run a bash script that includes df -k command on each server. The output text will be gathered for parsing and analysis. I'm having trouble understanding how to assess the df -k output correctly to arrive at the total disk space and usage.

For now, we are ignoring networked and LVM mapped storage (though I suspect they will be more involved and complicated than this situaion). I'll deal with these in the near future. For now, I'm having trouble understanding simple scenarios.

Scenario 1: I created an Oracle Linux 7.9 VM in Oracle Cloud with a default boot volume of 46GB. The df -h output returned the following:

Filesystem      Size  Used Avail Use% Mounted on
devtmpfs        7.6G     0  7.6G   0% /dev
tmpfs           7.6G     0  7.6G   0% /dev/shm
tmpfs           7.6G  8.7M  7.6G   1% /run
tmpfs           7.6G     0  7.6G   0% /sys/fs/cgroup
/dev/sda3        39G  2.8G   36G   8% /
/dev/sda1       200M  7.4M  193M   4% /boot/efi
tmpfs           1.6G     0  1.6G   0% /run/user/0
tmpfs           1.6G     0  1.6G   0% /run/user/994
tmpfs           1.6G     0  1.6G   0% /run/user/1000

Question 1: What consistent logic could I apply to calculate the total disk space and usage? In this case, I can see that (sda3 + one entry of 7.6GB tmpfs) will get me to 46GB, so should I ignore subsequent 7.6G tmpfs entries and all 1.6G tmpfs entries? Should I simply ignore all tmpfs entries, given that tmpfs is volatile and not real storage? In which case, how would I arrive at 46GB total storage?

Scenario 2: I created an Oracle Linux 7.9 VM in Oracle Cloud with a default boot volume of 200GB. The df -h output returned the following:

Filesystem      Size  Used Avail Use% Mounted on
devtmpfs         30G     0   30G   0% /dev
tmpfs            30G     0   30G   0% /dev/shm
tmpfs            30G  8.8M   30G   1% /run
tmpfs            30G     0   30G   0% /sys/fs/cgroup
/dev/sda3        39G  3.5G   35G   9% /
/dev/sda1       200M  7.4M  193M   4% /boot/efi
tmpfs           5.9G     0  5.9G   0% /run/user/0
tmpfs           5.9G     0  5.9G   0% /run/user/994
tmpfs           5.9G     0  5.9G   0% /run/user/1000

Question 2: This is more confusing. How would I arrive at 200GB total disk size? It appears I'd need to count sdaX AND (all tmpfs entries) to get to 200GB. I'm having trouble finding a consistent logic for both scenarios.

I hope my questions are clear. I'd be glad to provide any additional details and/or clarifications.

84
1 + 3
df
Michael Hampton avatar
cz flag
Michael Hampton
Have you tried running `fdisk`?
0
Reply
rc1 avatar
de flag
rc1
Just learned about fdisk based on the answer below. The only issue is that it's not compatible with Solaris (fdisk -l does not work), so we will need to figure out slightly different approaches for non-Linux environments. df -k returns consistent outputs across all Unix/Linux, but it's clearly insufficient on its own. Thanks.
0
Reply
rc1 avatar
de flag
rc1
Adding update for anyone else dealing with the same issue. I used "fdisk -l" on Linux (as suggested in the answer below), and "iostat -En" in Solaris 10/11 , followed by some careful parsing. Combined with df -k + (fdisk OR iostat) gives me pretty good coverage on overall disk size and utilization. I have not tested any of the above on AIX.
0
Reply
Score:3
berndbausch avatar Server
us flag
berndbausch

tmpfs filesystems are RAM disks and have nothing to do with your storage media.

In the second scenario, your 200GB disk contains a 39GB root filesystem, a 200MB boot filesystem, possibly swap space of unknown size and probably a large amont of free space.

To see swap space, run swapon -s. To see free space, use a partitioning tool like fdisk -l or parted /dev/sda print.

You don't seem to be using filesystems on LVM volumes or network storage. They would be listed by df.

0 + 2
rc1 avatar
de flag
rc1
Thanks for the clarification. So in short - if I understand this correctly - we cannot accurately determine the total disk size using just "df", and must incorporate "fdisk -l" in the data gathering and analysis? Or is there some practical estimation that can be done with just the df command? Thanks.
0
Reply
berndbausch avatar
us flag
berndbausch
Your understanding is correct. `df` only reports the space used in mounted filesystems. It does not report unmounted filesystems, partitions without filesystems or unpartitioned space.
0
Reply
ธงชาติไทย
Elon Musk
I sit in a Tesla and translated this thread with Ai:

mangohost

Post an answer

Most people don’t grasp that asking a lot of questions unlocks learning and improves interpersonal bonding. In Alison’s studies, for example, though people could accurately recall how many questions had been asked in their conversations, they didn’t intuit the link between questions and liking. Across four studies, in which participants were engaged in conversations themselves or read transcripts of others’ conversations, people tended not to realize that question asking would influence—or had influenced—the level of amity between the conversationalists.