Server reliability - UK Web Hosting | Dedicated Server Windows and Linux VPS Forum

**tonylees** · #1 (**permalink**) 14-07-2008, 11:48

My Private server died Sunday 13th July Early hours and was down until I discovered it at about 8:30am and NOW it has died again Monday 14th July... it doesn't look good when your customers call and ask if theres a problem and I can't tell them! Why can't Euk inform there clients of the problem?

I have been on live support for about an hour now and this is the longest it has taken to get the server back up and running..... This is not looking good

**eUKhost.com** · #2 (**permalink**) 14-07-2008, 13:40

Hi Tony,

We are trying to find out why the main hardware node got rebooted. Our system admins are going through all logs on the main hardware node of your VPS Hosting and they will get back to me with details once they manage to find out what went wrong.

**tonylees** · #3 (**permalink**) 14-07-2008, 16:45

The server was back online by about 2 o clock this afternoon, but it had been down from approx 10am.... isn't there anything in place to warn when a server is failing? so it has been down for approx four hours today and roughly 8 hours Sunday... 12 hours in July! surely that is more than is allowable under the 99.95% uptime guarantee and that isn't counting other times it has been out of action!!

I have lodged a ticket with Quality Control to see if I can get any answers.

**eUK-Scott** · #4 (**permalink**) 14-07-2008, 17:11

It turned out that the virtuozzo software was having a small bug itself with iptable modules which forced the main hardware node reboot. We have had to update this manually, and it's working fine now.

We have also discussed this with Parallels and they assured us that this will not happen again.

It is possible that your VPS Hosting was not set to autostart with node when node got rebooted. We have set autostart for all VPS Hosting's on the main hardware node.

**ServerOutage** · #5 (**permalink**) 14-07-2008, 19:27

So this 'little bug' with parallels will be common knowledge? Probably affecting all other parallels clients?

Is there any expanded information on their forums?

Quote:

It turned out that the virtuozzo software was having a small bug itself with iptable modules which forced the main hardware node reboot. We have had to update this manually, and it's working fine now.

I think this is a very poor explanation to give paying customers for such a lengthy outage.

As for having machines set to auto start surely this is fundamental? I've played with openVZ and that is like basic stuff.

**tonylees** · #6 (**permalink**) 14-07-2008, 20:41

Thanks for your input, I am not an experienced user and have no knowledege regarding the server and how it operates, is there anything I can do to help stop this happening again?

**eUKhost.com** · #7 (**permalink**) 14-07-2008, 21:26

Quote:

Originally Posted by tonylees

Thanks for your input, I am not an experienced user and have no knowledege regarding the server and how it operates, is there anything I can do to help stop this happening again?

Hi Tony,

This happened due to incorrect rules in iptables of main hardware node. There was nothing wrong on your VPS Hosting so you can't do anything to help us with such problem.

problems created by wrong iptables rules can't be detected easily as log files don't record anything for such problems.

**_Tux_** · #8 (**permalink**) 15-07-2008, 18:20

Quote:

problems created by wrong iptables rules can't be detected easily as log files don't record anything for such problems.

but surely you must have network monitoring software to check the availability of your servers and that would have alerted you immediately that iptables was causing an issue...

**Andi** · #9 (**permalink**) 15-07-2008, 18:28

Quote:

Originally Posted by tonylees

Why can't Euk inform there clients of the problem?

I agree, it would be nice to at least get an email from you guys stating that there is an issue as opposed to finding out from our clients.

**eUKhost.com** · #10 (**permalink**) 15-07-2008, 20:05

Quote:

Originally Posted by Andi

I agree, it would be nice to at least get an email from you guys stating that there is an issue as opposed to finding out from our clients.

Our Billing System has no such feature to through which we can notify customers hosted on a particular node. We are working on a new monitoring system which will notify each VPS Hosting / Semi-Dedicated / Dedicated Server customer if their server / VPS Hosting goes down.

We have ordered a new NAS Storage system which will be used to host /vz partition of all new Linux VPS Hosting nodes and D: of all new Windows nodes. This will help us to get rid of file system problems and disk read write problems are more frequent on VPS Hosting nodes and only way to get rid of those is to have a highly redundant NAS Storage.

I will talk to our developers and let you know how soon we can provide you with access to our new monitoring system.

**eUKhost.com** · #11 (**permalink**) 15-07-2008, 20:06

Quote:

Originally Posted by _Tux_

but surely you must have network monitoring software to check the availability of your servers and that would have alerted you immediately that iptables was causing an issue...

monitoring system will notify only if any specific service goes down. There's no such monitoring script which will notify reason for server reboot or shutdown.

**tonylees** · #12 (**permalink**) 15-07-2008, 20:10

Will it be possible for the system to notify when the server is back online? How will it notify account holders? Would be great by text

**_Tux_** · #13 (**permalink**) 15-07-2008, 20:20

Quote:

Originally Posted by eUKhost.com

monitoring system will notify only if any specific service goes down. There's no such monitoring script which will notify reason for server reboot or shutdown.

but at least we/you know that its down. you say the latest problem was because iptables did something odd, if you were actively monitoring the network availability of the server you would have easily seen that something was up and be on your way to fixing it before any of your customers noticed. (hint: we use a freeware app called "the dude" to monitor our internal servers at work)
instead you seen to have a policy of "everything is working fine until we actually fix the problem, then we can try say its the customer's fault. must be their internet connection or something. oh wait, something did break but we don't know why"

i'm now also quite concerned that you have ordered "a new nas system" thats "highly redundant". surely if you have a file system corruption problem, making that redundant isnt going to fix it. you'll just have redundant copies of corrupt data!!! (and a slow redundant copy if its on a nas as opposed to a fibre channel).

I also get the impression that the nas will hold data from multiple VPS Hosting nodes (i really hope i'm wrong on that) which i find very alarming from a reliability and performance point of view