Elasticsearch memory issue?

[catadetest (cata)]

Hi everyone,

As my manager told me we should expect more traffic in the following days, I would like to be prepared with the Elasticsearch servers performance.
So I did the following investigation:

1. Verified my load / memory usage by sysstat sar tool in the last 10 days. Everything is normal, load was low (between 0.1 and 1), memory consumption was low too (about 52%).

#load
20:35:01      runq-sz  plist-sz   ldavg-1   ldavg-5  ldavg-15   blocked
20:45:01            0       513      0.25      0.20      0.18         1
20:55:01            1       508      0.60      0.35      0.25         1
21:05:01            0       510      0.21      0.16      0.18         1
21:15:01            0       512      0.17      0.20      0.21         1
21:25:01            2       508      0.08      0.08      0.13         1

#memory
20:45:01      1185916  14868316  17358732     52.78    650760   7081224  17666272     53.72  21496064   3417884      1880
20:55:01      1167908  14868224  17359884     52.79    650768   7097812  17683856     53.77  21495788   3434084      1756
21:05:01      1188644  14867416  17359712     52.78    650772   7075932  17689732     53.79  21496076   3412632      1320
21:15:01      1195212  14867424  17359520     52.78    650788   7068988  17659912     53.70  21496040   3406192      1176
21:25:01      1193244  14867600  17359336     52.78    650788   7070608  17673440     53.74  21505852   3397340      1908
21:35:01      1286176  14865736  17360556     52.79    650788   6976356  17675876     53.75  21500104   3310292      1640

2. verified the Elasticsearch performance API, and this is where it gets interesting:
   It looks like the "ram.percent" value is 99% all the time. Is this a good thing, or it means I have to allocate more memory?

ip        name            heap.percent heap.current heap.max ram.percent ram.current ram.max node.role master cpu load_1m load_5m load_15m disk.used_percent disk.used disk.total
10.0.6.3  es03                      59        8.9gb   14.9gb          99        31gb  31.3gb mdi       -        4    0.62    0.58     0.56              7.80     9.7gb    125.4gb
10.0.6.2  es02                      64        9.6gb   14.9gb          99        31gb  31.3gb mdi       -        7    1.36    0.96     0.88              8.99    11.2gb    125.4gb
10.0.6.1  es01                      20        3.1gb   14.9gb          99      30.9gb  31.3gb mdi       *       19    2.20    2.18     1.93              7.89     9.8gb    125.4gb

Result of the "free -m" outputs the following:

root@es01:~# free -m
              total        used        free      shared  buff/cache   available
Mem:          32117       17156         423           1       14537       14497
Swap:             0           0           0

Also, there are no out of memory lines in Elasticsearch log file.
Would you consider that everything is normal here?
I'm just a bit concerned about the ram.percent value which is 99%.

Thanks!

[warkolm (Mark Walkom)]

@catadetest

It looks like the "ram.percent" value is 99% all the time. Is this a good thing, or it means I have to allocate more memory?

This is typical, as it's the OS caching commonly used files in memory to make access faster.
Most of those are likely Elasticsearch files, which is part of why Elasticsearch is fast.

It's nothing to worry about.

[catadetest (cata)]

Thank you for your time warkolm!