VPS and Performance

[martynh99]

This morning I was uploading a couple of small gif files to one of my website hosting on a Windows VPS Hosting when the VPS Hosting was shutdown.

Immediately got onto live support who tell me that my VPS Hosting was running 90% of the physical nodes CPU.

I was told that
- 'I think you are running very heavy applications on it'
- 'Might be MSSQL Server or ftp service creating load'

Now, there are no heavy apps running on there - there is 1 VBulleting instance running live and that is by no means heavy usage.
I have MSSQL installed but no databases conifigure yet.
I had just transferred about 4 or 5 gif files each of about 12k each - maybe that had overloaded FTP then

I have been sent a graph showing that my VPS Hosting was reporting 90% but of course that only comes from the VPS Hosting software - it doesn't look inside the VPS Hosting at whats going on in there.

Luckily for me, I am recording stats within the VPS Hosting and there was nowhere near 90% cpu utilisation, it was well below 20% and on further inspection a max spike of 12.6% was recorded about 30 minutes before the shutdown.

Overnight there were 2 small spikes and these tied in with the automaintenance jobs that run on the VPS Hosting nodes to send out stats.

I was also told that 'If your VPS Hosting is not having MSSQL Server or any other application then it is must having viruses' - this is after I have stated that inside the VPS Hosting node all is well.

When I suggested it was a bug in Virtuozzo I was told categorically 'It is not a bug. It is the software provided by swsoft who created virtuozzo'

Support have suggested that I monitor task manager inside my VPS Hosting to watch for problems, this despite me already having performance manager logging CPU/Mem/Bandwidth stats every 15 seconds. I only set these up because of problems I experienced when I first got the VPS Hosting - the console would appear to hang for no reason and this was with no website hostings running and the CPU was spiking then. Somebody from support 'tweaked' something and that fixed it - maybe the tweak neads tweaking a bit more.

Just one thing to add - I've held off transferring my more important sites over until I got a decent period of stability - I was just approaching that stage when this has happened so we're back to square one on this. My current feeling is go back to my original host.

[eUK-Nick]

I have attached a screenshot for your reference which clearly shows that your VPS Hosting was spiking. The image is from Virtuozzo which is the management software we use for our nodes.

These are images which cannot be tampered with.

[martynh99]

Originally Posted by eUK-Nick

I have attached a screenshot for your reference which clearly shows that your VPS Hosting was spiking. The image is from Virtuozzo which is the management software we use for our nodes.

These are images which cannot be tampered with.

Yep - I've seen the image, support emailed me. I've also sent in the 15 second sample csv file from Performance Manager to Support showing that the CPU INSIDE the VPS Hosting wasn't spiking.

In the CSV file are CPU/Avail Mem/Network Bandwidth/Page Faults_sec - as you'll see these stats do not reflect what the VPS Hosting software is reporting hence my suggestion that maybe the VPS Hosting software has a bug but from the Livechat session, SWSoft never have any bugs in their software !

I have just analysed the perf stats further and I can see out of the 2633 samples taken at 15s intervals only 5 were > 60% cpu and these were at 12:02am (3) 12:19am & 4:05am, all VPS Hosting maintenance times I believe for stats production and VPS Hosting domain reports to be mailed out.

[eUK-Nick]

Hello,

We have now forwarded your issue to SW-Soft. We are awaiting advise from them. We will update you once we hear from them.

[martynh99]

Originally Posted by eUK-Nick

Hello,

We have now forwarded your issue to SW-Soft. We are awaiting advise from them. We will update you once we hear from them.

Thanks,
Martyn

[martynh99]

Seems the network interface has changed aswell during this reboot.
Was called SWsoft Virtual Network Adapter
Now called SWsoft Virtual Network Adapter - Packet Scheduler Miniport

That means that during the reboot some other changes were made to the VPS Hosting config.

[mephisto]

Originally Posted by eUK-Nick

These are images which cannot be tampered with.

I know what you mean, but how else did you get the big white space there to cover up stuff??

[Rsync]

Originally Posted by mephisto

I know what you mean, but how else did you get the big white space there to cover up stuff??

The Other big white stuff is to cover up our other node details, as it's our confidencial data and We normally do not share network or core server information with our clients. We have had very bad experiences in the past where the information we give has been abused.

[mephisto]

Yes I know why it's there, I just found it ammusing that Nick said you cannot tamper the images, and then clearly had!

[eUK-Nick]

Originally Posted by mephisto

Yes I know why it's there, I just found it ammusing that Nick said you cannot tamper the images, and then clearly had!

I said we cannot tamper with the statistics which come out from Virtuozzo.

[martynh99]

Originally Posted by eUK-Nick

I said we cannot tamper with the statistics which come out from Virtuozzo.

No you definitely said images which cannot be tampered with


I'm sure the majority of us knew what you meant though !

Originally Posted by eUK-Nick

I have attached a screenshot for your reference which clearly shows that your VPS Hosting was spiking. The image is from Virtuozzo which is the management software we use for our nodes.

These are images which cannot be tampered with.

[martynh99]

Originally Posted by martynh99

Seems the network interface has changed aswell during this reboot.
Was called SWsoft Virtual Network Adapter
Now called SWsoft Virtual Network Adapter - Packet Scheduler Miniport

That means that during the reboot some other changes were made to the VPS Hosting config.

Would like to understand what went off here - clearly a config change was made.

Also just logged another performance case (copied you in on mail Nick)

[martynh99]

Originally Posted by eUK-Nick

Hello,

We have now forwarded your issue to SW-Soft. We are awaiting advise from them. We will update you once we hear from them.

Wow !

I really like their response from SW-Soft - interestingly I only got this supplied to me AFTER I re-opened the case with EUKHost that had been closed

The reply reads

We have received following reply from SW-Soft:
===
I logged in to your VPS Hosting and NW node and did not found any cpu load for the VPS Hosting at this moment.
===

What about checking and correlating the performance manager CSV log I sent in to the logs from the VPS Hosting/Physical Node or any recommendations on any additional logging that may be enabled to further help trap the issue should it occur again.

To me this sounds like no ticket was logged with SW-Soft - maybe I'm wrong but thats how it looks to me.

[eUK-Nick]

Hello Martyn,

SW-Soft was approached for help. We had indeed raised a ticket with them. This is being taken up further by us. I have replied to your email. Please allow some more time.

[vzAddict]

Hello,

There are some replies from SW-Soft which can't be given as it is to customers as they are as they contain some valuable information like Server Passwords etc. due to which we have sent you the main part of their reply. If you still want to get the complete reply, please re-open the Ticket we will update you.

[martynh99]

Originally Posted by vzAddict

Hello,

There are some replies from SW-Soft which can't be given as it is to customers as they are as they contain some valuable information like Server Passwords etc. due to which we have sent you the main part of their reply. If you still want to get the complete reply, please re-open the Ticket we will update you.

Yes, I also received the full reply ( may have been you who sent it me as it was from Mike K ? ) - it was equally as unimpressive

Following is the complete reply of SW-Soft:
===
I connected to your server via vzmc with password as ******** and followed to Top Resource Consuming VEs. Found that your VE (id 2413) takes only 30% of CPU and it is normal.
===

Reading from that then it would appear tht SW-Soft connected to the Physical Node (they have access ? ) and checked my VPS Hosting just the same way that EUKhost did - via the management console - As I keep asking in the ticket and in emails when is somebody going to look at the VPS Hosting logs against the perf logs I supplied from within the VPS Hosting and figure out why the management console said 90% and inside the VPS Hosting said way way less that that.

Remember the console said 90% average utilisation, but in my early posts on this thread I showed that out of the 2633 samples taken at 15s intervals only 5 were > 60% cpu. Over a 10 hour period that certainly does not equate to 90% average

Also nobody has explained why the Network adapter name changed after the VPS Hosting reboot - that implies to me that the software or config was updated




Further reminder, my initial 30 days is up tomorrow - i've not even transferred any of my customer sites across yet as I haven't had an acceptable period of stability as yet so it looks very likely i'll be putting in for my refund and taking my business away.