A number of websites were taken offline yesterday, and remained down this morning, after a load of servers run by Gloucester-based hosters Fasthosts fell over. Services on the hosting side are back up now, but the outage provoked angry responses. It was a faulty Dell switch that took the servers out, Fasthosts CEO Andy Burton …
dicky dell switch ?
didn't he present "world of sport" in the 70s and 80s ?
One switch took down a cluster
A certain lack of redundancy in their network design, methinks...
It depends. I've seen supposedly resilient networks taken out by a switch going apeshit. A particular favourite being where STP elections are being forced every second or so. As such no traffic is forwarded because the network is constantly listening and learning.
Second post and you're the first to suggest that a system you've never seen or heard of before has no redundancy.
One switch can take down a redundant system. The only thing that you can be sure of in a system is that the system will fail, everything else is merely a probability.
No, it can't
If you have a redundant design, one switch going down cannot take it down.
If one switch going down takes down your network, you do not have a redundant design, you do not have redundancy.
If you disagree, with the above, please explain why.
You can have an intended redundant design which turns out not to be.
You can have an intermittent failure which creates more traffic negotiating the failover than one switch can handle, making the system grind to a halt.
You can get a split brain scenario where both switches think the other is in charge.
These are off the top of my head... You could argue that you don't have a redundant system if it doesn't failover properly, but I'd counter that by saying you have a redundant system which didn't work.
I'm afraid you didn't read the question properly.
I said one switch GOING DOWN cannot take down a redundant design and if If one switch GOING DOWN takes down your network, you do not have a redundant design, you do not have redundancy, and asked for correction of this statement if you disagree.
You suggested scenarios where switches generate spurious network traffic, or split brain scenarios. Neither of which are a switch going down.
Fasthosts said "The switch tech that WENT DOWN comes from Dell; it disconnected the storage and the processors on some of the virtual servers. It was a specific cluster that was impacted."
Failure of a single switch disconnected storage and processors. There was no redundancy to account for for that switch failing. Therefore, it was not a redundant design.
They seem like a good idea - cheap, managed switches from the same supplier as your servers, your racks. Then you start putting traffic through them, relying upon them and then boom, its gone. Cisco kit is expensive, but Cisco kit having spent up on Dell kit first is even more expensive.
...I've seen Cisco switches completly twat a network as well....
Its possible for any vendors switch to have an issue, anything from human error in the code running the switch to something on the network doing something the software doesn't know how to handle can cause unexpected issues.
As its a Cluster running Vm's I believe its been made worse by the host having to restart all the machines again, its affected other hosts not just this one as well btw.
I've seen an entire company's UK operations taken down by a badly configured spanning tree, when an inadvertant user accidentally plugged two ports into each other when trying to connect a laptop in a meeting room. They were Nortel switches, but the effect it'd still happen on Cisco.
When will people learn
Fasthosts = 1&1 = Sedo = Shite
Hmm. I'm on FastHosts and 1&1, and I'm not going to disagree. The problem is that the information isn't readily available.
Mine wasnt down for long
Think i had outage for around 20 mins last night.
No-one can be as bad as Webfusion though for their VPS'. Mine used to go offline weekly, lost me a client. Apart from this hiccup I have been very happy with Fasthosts and will continue to use them.
salt in the wound!
I was one of their "handful" of customers who had their server in read only mode since Sunday night.
The certainly could of kept people better updated - updates were being posted ever few hours - and their twitter feed kept going with marketing tweets - the best two were promoting their "resilient virtual servers" and "The benefits of B2B blogs"
Tell them you're leaving and they'll reduce your hosting
I was reviewing my Fasthosts accounts the other week and decided to move one of them to another ISP as it was cheaper. I submitted the request to cancel the account and a week or so later... poor customer service... they phoned me, asked me why I was cancelling my account and offered to half the cost there and then if I stayed with them. So I did. I've not contacted yet though about reducing the costs of my other accounts.
The cost of hosting must be cheaper these days than it was 10 years ago yet there hasn't been much of a reduction in the cost of the service for customers. Both Fasthosts and Webfusion offer similar products and prices. A bit of a monopoly methinks.
switched off the wrong server
Ha. Not so long back, I contacted these clowns asking them to switch off one of the last two servers we had with them, gave them the unique server name + account number + pin.
Three weeks later, they duly shut down the other one.
When they put it back online, they then forgot to hook up their own backup systems again, so it was running without backups. Utterly amatuerish.
I had a fasthosts account once a few years ago. The service was so poor, I had to write a monitoring script to check the site availability every 5 mins and alert them when their server went down by emailing their support desk automatically. They werent very happy about that. AVOID.
Blackberry + Cisco Switch = 3 Day Outage
I used to work for Fasthosts, and I have to say, that when the shit hits the fan technically, they really do try to resolve it as quickly as possible.
I know they have a lot of bad press, but from someone that knows the inside scoop, they will have been pulling every resource available to sort the issue out.
Let the downvotes commence!.... :-/
The firmware that Fasthosts applied isnt just a firmware patch. If they are using the Dell 24 port 10Gig switches (rebranded broadcoms from memory) then the new firmware is a full code rewrite.
Re the questions of redundancy, I gather Fasthosts issues were around latency between the VMware hosts and the Storage. This caused weird issues. Having two switches wouldn't help, as technically nothing failed....
Webhosting company has outage. Not major news!
Can only but wonder how many of those 127 complaints were from re-sellers?
When I saw the displays on the end wall of the coffee room in the the top floor attic where front line support is located showing the stats for re-sellers signed up that day I knew this was not a firm I wanted to work for and turned them down flat when they offered me a role.
Ah... that's why they called me yesterday
Only just twigged. A marketing guy called me yesterday to find out if I was happy with them. I've had no outage on a Linux DS, unless I managed to miss it.
I've been with them about a month (don't ask), and my major problem at the moment is that I can't talk to any other FastHosts users. It turns out that they closed down their forums a few years ago, and haven't re-opened them. Any forums out there? I'm on webhostchat.co.uk occasionally. Please don't say twitter. Please.
Good old Slowhosts...
Wouldn't touch them with a barge pole, having wasted hours and hours of time trying to do basic joomla installs with different clients who use them; the same installs all done and dusted within an hour with 1&1, who are no angels, but at least their basic services work at acceptable speeds.
- Mounties always get their man: Heartbleed 'hacker', 19, CUFFED
- Batten down the hatches, Ubuntu 14.04 LTS due in TWO DAYS
- Samsung Galaxy S5 fingerprint scanner hacked in just 4 DAYS
- Feast your PUNY eyes on highest resolution phone display EVER
- AMD demos 'Berlin' Opteron, world's first heterogeneous system architecture server chip