GPLHost 13+ Hours Outage in 2 Days

Tagged in

There are some very good Xen hosts. SliceHost being one that I have had experience with — it has rock solid stability, very good uptime, great performance and all the flexibility of Xen at an incredibly price. Too bad they are not in Australia. I don’t know how bad a Xen VPS host can be, but I certainly have some bad experience with GPLHost, whom I have reviewed their Sydney-based VPS 6 months ago. They were “tolerable” for the first few months, but got worse over the last couple of weeks.

There are some issues like

  • Regular downtime. I am having difficulties keeping the VPS up for more than 10 days.
  • IO wait. Sometimes it is so severe that an ls command will lock up the VPS for a few seconds.

These issues certainly made running a production website extremely difficult as you have visitors accessing your data 24/7! Last two days had also been quite dramatic.

Sunday 3 June at 3:50am
Server down
Sunday 3 June at 11:15am
Server up — offline for 7 hours and 25 minutes
Monday 4 June at 3:10pm
Server down
Monday 4 June at 8:50pm
Server up — offline for 5 hours and 40 minutes

Take today for example, the server was still running, however the entire thing was just not responding — can’t return web pages, can’t access SSH, cron does not run, etc. Fortunately it still somehow respond to SNMP requests, so I had grabbed some graphs.

Network at GPLHost VPS

As you can see, the outbound traffic dropped to 0 after around 3pm, although it is still getting inbound traffic. Now let us check out the load on that box.

Load at GPLHost VPS

Normally my load is less than 1, but it has raised all the way to 30+ just before the reboot — something is blocking the processes from running, and I have no idea! Pain of running a VPS.

Yes it sucks. Consider thousands of dollars lost due to this server outage! Oh wait, I made that previous one up, and it was probably in the scale of a few dollars, but still — extensive downtime is annoying. Worse, as it has already happened two days in a row, I am now pretty scared that it will happen again during this week, leaving my websites inaccessible for more hours.

While GPLHost usually has pretty good support, but no one was there handling my support email today. Even if it has been actively worked on, I am not sure whether it is going to be the saving grace this time. Not sure what secret sauce other Xen providers prepare their servers with, but I think GPLHost certainly needs some of those…

However, hunting for another good VPS provider in Australia remains difficult. Vigabyte looks like is having some hot technology, but our friend Nicholas went and tested, and couldn’t even get the VPS he paid for after 11 days. Dodgy.

Currently the cheapest option seems to be getting a dedicated server, install OpenVZ or Xen, and then share the cost with friends. Servers Australia have some very cheap dedicated servers (AUD$99/month for a AMD Semperon 3400, 1GB RAM, 160GB SATA HDD, 100GB/month transfer). But as the “Order Now” link does not even bring you to the right area, I think I better go and check my dodgy metre on this host first.

Comments

Gravatar

I thought you might have found the dedi server company I was looking for with that price, but its not :(

I found a pretty sweet deal for dedicated servers in Australia but I didn’t bookmark the site and I can’t find it again!! It’s pretty annoying since I’d probably sign up… At the moment I’m signing up to a few different places, its a bit pricey but oh well, do what you need to do ti get it all working properly. But its not like they are just going to be sitting there doing nothing I have uses for each one, I just don’t have enough time to set it all up …

I’ve just got my slice512 for 6 months today, so will need to set that up soon… I have bought a 2nd VPS from VPSLand that hasn’t been setup, had that for a month and a half so far…

maybe I got out just intime from GPL Host :P As I mentioned to you I’ve moved node and seem to be doing all right now. Fingers crossed it stays this way.

Gravatar

Hi Nicholas,

I think I probably won’t dismiss Servers Australia that quickly due to their sign up pages as I did hear some good reports about them, but I’ll say $99 for a 100GB dedicated server in Sydney is really “too cheap”. As of GPLHost, Thomas emailed me today saying they might migrate me onto other servers. We’ll see.

As of setting up the VPS, I guess that’s the pain when you want to use Gentoo. “emerge —sync && emerge world” after getting the root prompt, and then come back two days later to install apps :)

Gravatar

Yeah emerge —sync && emerge world easy, I mean the next step. Like the services that run on it, mail, apache, dns, how it all works with existing services I have on other boxes, etc :) I’ve finally worked out a way to have multiple sites with multiple owners and separate everything so each site is its own and it can be moved from host to host etc

This stuff is all simple on its own, ie a website, step, step, step, done. but if you want to exist in the infrastructure of this then it all needs to planned out and bits put in place.

Gravatar

Might have some resolution on the Vigabyte situation http://www.nicholasorr.com/2007/08/03/vigabytesmartyhost-update/

Gravatar

Hi,

We had some big hardware problems with that server, that the collocation partner we are using recommended to us. Since the beginning, we experienced many problems with the HDD controler, and even after asking and asking again, nobody either at ASUS or the hardware supplier replied to us that the problem would be fixed.

We were promised that a KVM over IP would be setup on that server that costs us a hell of a lot. This has never work either.

So we have bought a new server with the type of hardware (Supermicro PDSMI+ motherboard with Supermicro casing). All of the VMs are now migrated to it, and we believe that the issues are over now.

We also added smartd monitoring to all of our servers, so it now reports errors BEFORE it actually kills everything, so we can replace our HDDs on time.

Last thing: I really believe that you went on one of our worst server, and that this experience is not representative of our service. Let me show you this:

zigo@node0105:~$ uptime

07:28:10 up 710 days, 11:56, 1 user, load average: 0.23, 0.16, 0.15

try to beat it ! :)

Thomas

Gravatar

Thomas,

Yes the new Core 2 Duo server that I moved to has been rock solid so far. Many things that I have been able to do before (compiling heavy packages for example) are now handled with ease. Yes it sucks when the hardware is having issues :(

Big thanks for getting us migrated.

Post new comment

The content of this field is kept private and will not be shown publicly.

More information about formatting options