Troubleshooting repeated server crashes

[Peter Downes]

Hi Forum

My server has crashed 5 times since 31/03/2016 - Plesk and websites unavailable.

My support has so far reported

- "All the server hardware is fine". When the hardware checks completed, they noted that all hardware had passed, although when booting up it was using a previous Linux kernel and showed an error 24.

- My "servers CPU is being utilised due to there not being sufficient memory to store processes in memory". They believe that most likely "a zombie script is running and not closing down".

- Spam Assassin disabled
- Global Maximum memory for PHP scripts reduced from 128mb to 32mb
- My external reboot button has not available for a week or so

I am concerned we are not getting to the bottom of the problem.

I would appreciate some advice on
- what I might be able to do to help.
- the right questions to be asking support.

 

[Lloyd_mcse]

Hi Peter Downes,
I suggest you have a look through the logs, you can find a list of relevent logs here...

Plesk for Linux services logs and configuration files

And of course you have tools like top that will list running/sleeping/zombie processes.

I hope that helps.
Regards

Lloyd

 

[UFHH01]

Hi Peter Downes,

to investigate issues/problems/errors/failures on your server, it needs far more informations than you provide, Peter Downes.

We have no idea, what server you use, with which specifications.
We have no idea, what OS you use.
We have no idea, how you updated/upgraded/patched your OS lately.
We have no idea, what Plesk version ( incl. MU ) you use.
You don't provide any errors from log - files, apart from the information about your kernel error at boot time.

I am concerned we are not getting to the bottom of the problem.

Sure not... without any decent informations, there is only the possibilty to guess causes - we are as lost as you are and the possibilities for any cause is HUUUUGE! Just imaging, that you call your car mechanic, just telling him: "My car is broken" - you wouldn't really expect a satisfying answer to that question, how to repair your car, do you?

Yes, you will now answer next: "But I only asked for advice and to name the right questions for my support"... and the true answer from people willing to help you honestly should be: Please hire an experienced linux administrator, because your described issue isn't just a a quick "question & answer" - situation, it needs deeper investigations directly on your server. Please don't get this wrong, but this forum is a "Plesk-related" - forum.

 

[Peter Downes]

Thank you @UFHHO1. I am always grateful for your great advice.

Please hire an experienced linux administrator

You will be pleased to hear this is a managed server with 24/7 support from professionals.

Thank you @Lloyd_mcse. I'll follow that up.

 

[Jesse Fitzgerald]

Hi Peter. I'm sorry you are having trouble. Have you looked through /var/log/dmesg and /var/log/messages for evidence of a kernel panic? Which distro are you running? ...and also, What brand and model is your server? --Jesse

 

[Peter Downes]

Thanks for your kind interest Jesse. Sorry for the late response - I have been up to my neck in it.

Please don't get this wrong, but this forum is a "Plesk-related" - forum.

That's a really good point that I had kind of forgotten - everybody has been so helpful that I just leapt right on the board in my panic / frustration.

There is no problem with Plesk on the server. I'm sure its deeper than that - I think due to a lack of memory even though plesk server health reports all OK.

--
Real memory usage: 27.6% used (2.12 GB of 7.68 GB) 1.98
Swap usage: 1.7% used (133 MB of 7.81 GB) (?)
Swapping throughput: 0.25 pages 20.29
--

So I will try to keep the support team on the case until they find problem.

 

[Peter Downes]

Just discovered there is a problem with Plesk.

My Plesk admin health monitor displays: Real memory usage 34.4% used (2.64 GB of 7.68 GB)
A command line check shows: Mem: 8050732k total, 7743788k used, 306944k free, (That's 96% used)

Is a health monitor that is wrong more dangerous than no health monitor at all?

 

[Peter Downes]

Ah sorry. Just for the record - we misinterpreted the CLI readings. Plesk health monitor readings were fine.