Welcome to the High Availability Hosting Blog

This week we have ordered an additional fibre into the Sheffield Data Centre.

The fibre will be provided by BT for connection direct into their 21CN network, with onward Internet transit.  This brings our fibre connectivity N+1 in line with power and cooling which is already redundant.  BT also brings a second transit provider to us.

We’re all really pleased here at HA Hosting, as another fibre will enable us to trial several in-the-pipeline products, including the possible return to dedicated gaming colo.  The additional bandwidth has already been partially pre-sold to existing customers requiring extra capacity, and will allow us to meet future needs.

Due sometime in April, we’ll of course blog the install when it happens, and if BT have top open any more manholes, we’ll try to get pictures  :-)

We’ve today ordered 4 x Intel i7 servers to host Virtual Private Servers at the Sheffield Data Centre.

These servers will suppliment our existing AMD Phenom II server cluster, giving customers either AMD or Intel options.  We’ll be doing some testing this week before bringing the servers into live service next week.

Prices to customers won’t change, and we may run a special offer in February for Intel based VPS’s.

HA Hosting provide Virtual Private Servers from just £10 per month, with a choice of Windows or Linux Operating Systems.  See http://www.sheffielddatacentre.com/servers/default.aspx for full details.

The last customers have been migrated from Exchange 2007 to our new Hosted Exchange 2010 platform.

The legacy (3 year old) Exchange, Management, Domain Controller servers will now be soft-decommed before being switched off at the end of the month.

The website still says Exchange 2007 – this will be fixed!

Hosted Exchange 2010 10GB mailboxes are £6.00 per month.  Unlimited and ” basic” mailboxes are also available, contact us for more information, a trial, or to order.

We’ve started carrying out power audit work at the Sheffield Data Centre, accurately documenting what each customer & device is allocated, and what it’s using.

All co-lo customers will be contacted over the weekend with regards to what they’re allowed, and if they’re using any excess, but for the sake of completeness here are the allowances (unless  by prior arrangement):

—

Single Server

• 1 x power socket
• 1A power

Quarter Rack

• 4 x power sockets
• 2A power

Half Rack

• 6 x power sockets
• 5A power

Full Rack

• 12 x power sockets
• 10A power

Additional Resources

• Additional power @ £22 per month per Ampre
• Additional sockets @ £2.50 per month

—

All prices exclude VAT.

Customers are not permitted to plug devices direct into the UPS, only into PDU’s provided by us.

Customers may provide their own additional PDU’s, but they should plug into our PDU.  We use our PDU as the measuring point for power consumed.

—

Our wholesale energy prices have increased this year, we’ve decided not to pass on this cost to customers at this time, however we will be invoicing customers more activly with regards to their actual power and socket usage.

Last night our monitoring system alerted us to 4 BGP resets on our edge router to Virgin Media.

The first 3 resets lasted less than 2 minutes, the 4th lasted 11 minutes.  By the time the on-call engineer logged in at 02:10 everything was back to normal but we did get lots of log data to throw at Virgin Media.

Unsurprisingly, pi$$ed of customers contacted us this morning, we’re sorry you experienced an issue out of our control.

We (HA) tell customers when we’re performing scheduled maintenance, a rough time we’ll start, and a best-estimate of the duration.  It’s a shame not all suppliers do this.   We’re still waiting for an official statement from Virgin Media.

Whilst investigating a support request from a customer yesterday, we discovered some Mozilla Thunderbird users had problems sending email through our new Basic Mail service, powered by MailEnable.

If you use our web hosting email, or Basic Mail services, and have problems sending email (receiving is fine), try the following links for guidance on SMTP Authentication:

http://www.mailenable.com/Help/Files/configuringthunderbird.htm

http://kb.mozillazine.org/SMTP_authentication

Outlook users seems to be OK, as Outlook uses a different method for SMTP Authentication (by using the same credentials as POP3).

This information may be relevant to all customers who have been migrated to our new Basic Mail email service (also used for web hosting mailboxes) over the last weekend, or new customers using Basic Mail and having problems with Thunderbird.

Thanks to Sara for her help testing the fix!

We’ve had a large number of customer VM’s to provision over the last week, which has caused delays to customers who ordered VM’s Thursday PM and Friday (22nd / 23rd).

We’re clearing the backlog today, so all VM’s ordered before end of play yesterday will be provisioned today.

Sorry for the delays!

This morning between 05:30 and 06:40 (along with the power redundancy RFC), we completed the revised configuration of Spanning Tree on the colo switch stack.

The RFC involved configuring a new Root Bridge, and creating a second Trunk between two switches for added redundancy.  Everything is behaving as expected.

Clients – If you suspect any network “oddities” due to the new Spanning Tree config, open a ticket with us and we’ll check it out.

This morning between 05:30 and 06:40 we completed power work on the colo switch stack at the Sheffield Data Centre.

The work allows the feed from 2 x UPS’s to power the switches and routers, with auto-failover and manual bypass for planned maintenance.  During the work we paused some VM’s incase of cluster errors (but there were none), with everything being brought back online cleanly.

We’ve scheduled maintenance on our colocation network stack this Monday morning (19/09/2011) between 05:30 and 06:30.

The work is to provide power-redundancy and updated Spanning Tree settings to our colocation switch stack and core routers at the Sheffield Data Centre.

ALL SERVICES WILL BE DISRUPTED for around 3 minutes whilst the edge routers reload.

We are also disabling Fast Link on colo switches, meaning each colo switch port will reset for around 30 seconds after the change has been made.  Customers with dual/multiple links shouldn’t see a problem.

This is the work which was postponed from Friday 9th September.

iSCSI / SAN customers are NOT AFFECTED because those switches are already connected to newly configured switches.

As usual, if you would like to discuss this maintenance with us, give us a call on 0114 228022 or email “support @ hahosting.com” (without the spaces).

UPDATE 19/09/2011 @ 07:00 – All work completed between 05:30 and 06:40.