How can i optimize my server?

E

ericosman

Verified User
Joined
Nov 25, 2019
Messages
627
Location
Weert
Hi all,

For a while now i'm running my servers, but i think they are not running optimal.

1. server info:
[CONTABO L SERVER]
AlmaLinux 8
AMD EPYC™ 7282
8 V-cores
30GB RAM
400GB NVME storage
Click to expand...

Could upgrade to :
[CONTABO 5 SERVER]
AlmaLinux 8
AMD EPYC™ 7282
16 V-cores
64GB RAM
900GB NVME storage
Click to expand...

if that would make a good change, why not :)

2. PHP info:
php1_mode
php-fpm
php1_release
7.4
php2_mode
php-fpm
php2_release
8.0
php3_mode
php-fpm
php3_release
7.3
php4_mode
php-fpm
php4_release
8.3
php_ini
no
php_ini_type
production
php_timezone

secure_php
yes
x_mail_header
yes
Click to expand...
3. MYSQL info:
mariadb
10.11
mysql
8.0
mysql_backup
yes
mysql_backup_dir

mysql_backup_gzip
no
mysql_force_compile
no
mysql_inst
mariadb
Click to expand...
4. WEB Server Settings

apache_mpm
auto
harden_symlinks_patch
yes
http_methods

litespeed_serialno

mod_ruid2
Select
modsecurity
no
modsecurity_ruleset
owasp
redirect_host

redirect_host_https
no
unit
yes
use_hostname_for_alias
no
userdir_access
yes
webserver
nginx_apache
Click to expand...

5. what i did :
update /etc/my.cnf
[mysqld]
max_allowed_packet=64M
local-infile=0
innodb_file_per_table=ON
#tmp_buffer_size=64M
max_heap_table_size=64M
table_definition_cache=5120
performance_schema=ON
key_buffer_size=32M
innodb_buffer_pool_size=1024M
innodb_log_file_size=256M
innodb_log_buffer_size=32M
tmp_table_size=32M
key_buffer_size=6M
Click to expand...

I have read that there are ways to optimise settings for php etc, but i could not find out what would help me.
I do have REDIS activated, but i think it's not set up too well.

Thank you for all the help in advance!
 
Last edited:
ericosman said:
but i think they are not running optimal.
Click to expand...
What make you think is not ? any benchmark ?
ericosman said:
Could upgrade to : Contabo 5 Server
Click to expand...

Yes Contabo give lot of resources but single core score is not really high, I am not sure how the conversion is made from an real core
for example how many vCPU are they "generating" from an physical core?
Start with TOP command in your SSH and see the "steal" values , if they are above 10 then your server is oversold

If you want i can pass you some test scripts , you can also use https://hetrixtools.com/ to monitor your server resources (its free for 15 monitors/servers)
 
Last edited:
Active8 said:
What make you think is not ? any benchmark ?
Click to expand...
Well, clients of mine are hosting wordpress sites (yes i know.... wordpress.. haha)
And the page speeds all give (even after i optimize the site, eg cache compression etc.)

"Reduce initial server response time"


Active8 said:
Start with TOP command in your SSH and see the "steal" values , if they are above 10 then your server is oversold
Click to expand...
https://i.ibb.co/k4bhnWM/Scherm-afbeelding-2024-01-20-om-16-44-52.png i guess that looks good (?)

Active8 said:
If you want i can pass you some test scripts , you can also use https://hetrixtools.com/ to monitor your server resources (its free for 15 monitors/servers)
Click to expand...
Well, please if you have some scripts that i could test run, i'd be happy to do so.
 
ericosman said:
"Reduce initial server response time"
Click to expand...
That could well be also caused by Contabo. We had several VPS systems there. And especially one which was fast in the beginning became very very slow while not more customers were added.
Also another one with less stuff on it became slow.
Example. When I rsync to another server, it takes maybe 1 or 2 secs and it's connected. At a certain point a VPS which we used as backup with only a few customers on it and not much visited, took around 10-15 seconds before the connection was made. And then sometimes during the night it was again 1-2 seconds. Also sites started to go slow.

Contab has very good prices, maybe too good to be possible, so it might be they are overselling a bit or reducing things the get more clients on a vps. I don't know, but just wanted to point out it doesn't have to be your vps systems as (only) cause.
 
Declared CPU not very fast itself per single thread (as php and mysql processed). If it overselled - it become worse. Check as @Richard G advised you is there a lot of resources stolen, better to check when max loaded middle of day or near.
 
@Active8 depends on your needs - sometimes better to pay little more and have guaranteed resources.
like ex44 would be at least twice faster than 8 V-cores of overselled AMD EPYC™ 7282

Dedicated Server Hosting

Find your reliable dream server, optimized for high capacity use, ever-growing storage needs or rapid and parallel processing.
www.hetzner.com www.hetzner.com
 
Zhenyapan said:
@Active8 depends on your needs - sometimes better to pay little more and have guaranteed resources.
like ex44 would be at least twice faster than 8 V-cores of overselled AMD EPYC™ 7282

Dedicated Server Hosting

Find your reliable dream server, optimized for high capacity use, ever-growing storage needs or rapid and parallel processing.
www.hetzner.com www.hetzner.com
Click to expand...
They also have a server auction, where you can get even cheaper offers from previous customers
https://www.hetzner.com/sb
 
Thanks!

We also got several servers by Hetzner by the way. Very happy about them.
Some we got from the auction, you can make very good deals there indeed.
 
Am i wrong for thinking the following;
I have 2 VPS systems now at Contabo :
[CONTABO L SERVER]
AlmaLinux 8
AMD EPYC™ 7282
8 V-cores
30GB RAM
400GB NVME storage
Click to expand...

And i would be able to take the following system (for even less money):

Intel® Core™ i5-13500
incl. Hyper-Threading Technology
RAM:64 GB DDR4
optional max. 128 GB DDR4 (for additional charge)
Disk:2 x 512 GB NVMe SSD (Gen4)
(Software-RAID 1)
Connection:1 GBit/s-Port
Bandwidth guaranteed:1 GBit/s
Backup Space:100 GB
Traffic:Unlimited *
Available Locations


So than i would save money and have a better system, the only downside is that i will have to move everything over.
 
Well, we did the same. You've got a very fast system. We have an i7-6700 now for cheap.
I don't know if it was because it was busy but we were with normal ssd disks around 100mb/s slower than your system.
NVMe is faster.

I wouldn't think twice if I were you. Moving everything over is not that much work if you have enough with 512 GB disks.
 
ericosman said:
Would it not affect mail discoverability?
Click to expand...
Discoverability no, all ip's are fast to be found. There might be some time required to sync the new ip's with the nameservers yes.
And ofcourse, one never knows with the ip's, reputation has to be rebuild again, but normally it's not a big problem.

Or do you mean something else with "discoverability"?

However, be aware that they are using raid 1 and disk/ssd/nvme performance seems slower with raid 1.
You ofcourse can choose to not use raid 1 but that is your choice/risk.
I doubt things will go slower in your case with that fast cpu. :)
 
Well, the decision is made, bye bye Contabo... hello Hetzner, the only thing i'm waiting for now is them to respond to my request to unlock the mail ports before the end of the first month..

And a cheap CLOUD VPS 1 at contabo as a secondary dns (directslave) *if i can understand how to install it haha*
Or PowerDNS
 
Last edited:
ericosman said:
my request to unlock the mail ports before the end of the first month..
Click to expand...
Oh is that also for servers? I thought they only had this in place for their VPS systems, but I can be mistaken. We are long customer there so for that reason they don't block our new servers anyway.

PowerDNS... pfff.... I tried on a VPS but I really don't get it, found Directslave way easier so I'm running a Contabo VPS with Directslave as 2nd nameserver for 2 DA servers at the moment.
Working fine. :)
So I could help you with that if you want.
 
You must log in or register to reply here.
Back
Top