Question Ram usage 98% after 24 hours

[ChrisCP]

Hello everyone. Since the update from onyx to obsidian, the RAM usage is constantly full and SWAP is being used. Onyx's RAM utilization was just 24% and the server ran for days without problems. Nothing was changed to the domains or settings in Plesk. Does anyone have an idea why the RAM is full within 24 hours?

 

[Bitpalast]

How much RAM do you have? Does it fulfill the minimum system requirements for Obsidian (512 MB + 1 GB Swap)?

 

[ChrisCP]

16GB RAM and 19GB SWAP. After restarting the server, a few hours the RAM is filled with 60%.

 

[Bitpalast]

It is absolutely o.k. and normal that the RAM is filled. It could even be filled up to a 100%. This is the normal Linux behavior. RAM will be freed when processes require it. The only thing that seems to be wrong is the swap usage. It could be o.k. if the usage is minimal, but it seems that the system is using swap extensively? Have you checked what processes consume most RAM?
# ps aux

 

[AYamshanov]

Hi ChrisCP,

What OS do you use? Could you please also describe how you monitor memory usage in Onyx and Obsidian?

 

[tkalfaoglu]

If Linux, try the top utility.. Then hit M for memory summary.. My bet is on too many httpd processes or mysql.. the two big hitters on my machine..
Also consider adding more RAM. My server has 64GB and did not notice any changes in RAM usage.

 

[ChrisCP]

Hello to all. I use CentOS 7.7 with the latest updates installed. I always used the simple Server Information in Plesk UI under Onyx. After the upgrade to Obsidian i use htop. Htop shows me a lot of mysqld processes running. I use Mariadb 10.3.20. I think also that the mysqld eating my ram.
The question is, why in Obsidian and not in Onyx?

 

[Bitpalast]

In your /etc/my.cnf file or files included therein, what are your RAM parameters?
For example what is the innodb-buffer-pool-size set to?

 

[ChrisCP]

In my.cnf there is no entry. In information_schema.global_variables the entry for the innodb-buffer-pool-size is 134217728.

 

[tkalfaoglu]

Perhaps this would help:
How to Optimize MySQL Performance Using MySQLTuner

 

[Bitpalast]

How much RAM does MariaDB really use? So far, the only "real" issue described in this thread is a frequent swap usage. Could you please give more information on the real RAM usage and swap usage?

 

[AYamshanov]

May it depends on using different tools for diagnostic ("Server Information" on Onyx and "htop" on Obsidian)?

Had you used a Plesk Health Monitoring component until the upgrade from Onyx to Obsidian? If yes, I think you can see a history of memory usage changes by processes (database, mail, web) and system in total with Advanced Monitoring. The same about using the swap. Is it changed? Health Monitoring and Advanced Monitoring use the same data collecting service.

 

[ChrisCP]

"Real" use of the complete RAM is about 1,5GB, the rest ist chached and buffered and Swap increases... No i didn't use Health Monitoring because everthing was ok. For the time i have to restart the server every 1 1/2 days because swap slows down my server.

 

[AYamshanov]

It is strange if required to reboot a server every 1.5 days.

You can find "TOP15" processes which uses SWAP with the next command:

for file in /proc/*/status ; do awk '/VmSwap|Name/{printf $2 " " $3}END{ print ""}' $file; done | sort -k 2 -n -r | head -n 15

On my test tiny server, I see the next output:

# for file in /proc/*/status ; do awk '/VmSwap|Name/{printf $2 " " $3}END{ print ""}' $file; done | sort -k 2 -n -r | head -n 15
dockerd 25928 kB
mysqld 15564 kB
containerd 14440 kB
grafana-server 13920 kB
named 11996 kB
sw-engine 10628 kB
influxd 8192 kB
unattended-upgr 7984 kB
networkd-dispat 7756 kB
php-fpm 7356 kB
snapd 6052 kB
sw-engine-fpm 5068 kB
sw-engine-fpm 4080 kB
fail2ban-server 3992 kB
sw-engine-fpm 3324 kB
#

...and it can be ok if it isn't causing huge traffic from memory to swap and from swap to memory (swapping).

---

I also check the Advanced Monitoring graphs about memory. for the last 30 days. The last reboot was 18 days ago because of some OS updates.


Do you see a lot of read/write swapping operation for your server?
Could you please describe how you measure the performance of the server/website?

 

[Ales]

Just to note, swap can be flushed without rebooting. Not that it generally should be, linux is quite good at managing resources on its own...