Issue Server is down

[alegsa]

Hi. Sorry my English. I have a dedicated server with Plesk Obsidian. The server is going down from time to time and I can't figure out what is happening.

This is the chart of Advanced Monitoring of Plesk: a huge use of memory from Apache!:


An increase of sleeping processes too:


I have to reboot it usually from GoDaddy panel, because I can't even connect through SSH sometimes. When I reboot the server, it is back to normal.
The problem happens randomly (sometimes 12 hours, sometimes 1 week), it is not connected to scheduled tasks.

This is TOP when the server is with problems (you can see a lot of sleeping processes) :

This is TOP when the server is normal:


I changed from Onyx to Obsidian 1 month and a half ago. This problem is happening like 2 weeks ago.
I don't know where to start to collect data to figure out what is wrong. Maybe logging processes?

Thanks

 

[Monty]

Your 2nd screenshot of "top" shows that your server is out of memory (although you said that this is the screenshot when the server is working fine). But you have no swap space configured and all of your 4GB of RAM are used.
Once all available memory is used, your server will become dead-slow and (if activated) the out-of-memory killer from the kernel starts killing processes.

So you should investigate why your RAM usage is so high. You may have a website that is under attack and thus consumes all your memory.

Some starting points:

How to diagnose a DoS/DDoS attack and find websites under attack on a Plesk server

Applicable to: Plesk for Linux Plesk for Windows Question How to diagnose a DoS/DDoS attack and find websites under attack on a Plesk server? Answer On Linux For real-time attack Conne...

support.plesk.com

Resolved - memory usage by web service

I just can't explain myself this: from about 7.00 pm (when we no longer work from our web backoffice) until about 6 am, the Apache memory grows a lot, sometimes up to 70-80% The process that consumes the most is php-fpm: pool crm.100asa.it (and usually there are up to 5) Here is the chart...

talk.plesk.com

 

[alegsa]

Your 2nd screenshot of "top" shows that your server is out of memory (although you said that this is the screenshot when the server is working fine). But you have no swap space configured and all of your 4GB of RAM are used.
Once all available memory is used, your server will become dead-slow and (if activated) the out-of-memory killer from the kernel starts killing processes.
So you should investigate why your RAM usage is so high. You may have a website that is under attack and thus consumes all your memory.

I'm not out of RAM memory when the server is normal, the memory is CACHED:


see: What is difference between cached memory and used memory?

Some starting points:

How to diagnose a DoS/DDoS attack and find websites under attack on a Plesk server

Applicable to: Plesk for Linux Plesk for Windows Question How to diagnose a DoS/DDoS attack and find websites under attack on a Plesk server? Answer On Linux For real-time attack Conne...

support.plesk.com

Resolved - memory usage by web service

I just can't explain myself this: from about 7.00 pm (when we no longer work from our web backoffice) until about 6 am, the Apache memory grows a lot, sometimes up to 70-80% The process that consumes the most is php-fpm: pool crm.100asa.it (and usually there are up to 5) Here is the chart...

talk.plesk.com

No, it is not a DDoS attack. I have DDoS attacks in the past and I resolve them. I have check the websites logs and everything is fine. Besides the problem disappear immediately when I reboot the server.