Afrihost service - hosting complaint

[Sinbad]

NO

You can rebuild a mirror without reading the original disk?

 

[SilverNodashi]

3 disk raid 5 has two data segments and one parity segment for each stripe, distributed among the disks. Loss of one disk allows on the fly but slower reading of data. The set can be reconstructed when the failed disk is replaced.
As for your 1tb comment - the size of the disk is irrelevant.

With Raid 5 you need a minimum of 3 drives and will lose 1 drive space for redundancy. If you use 3 x 1 TB drive in raid 5 you will have 2TB volume to save data, one drive can fail and you can still read the data on the volume.

I would love to see you take out 1 drive (simulate a total drive failure) in a 3-drive RAID5 set, and still rebuild the RAID set. But, do this on a busy server with 3x 1TB HDD's. Forget about the nice theory about how RAID5 works for a second. Those RAID levels were designed with 36GB and 100GB HDD's in mind, many years ago. The rebuild will take about 18 hours, to two days, or more depending on the server's load. Chances are you're customer's will experience downtime as the server will take a performance hit. Even with a hardware RAID card, with BBU cache.
The next problem you sit with, is that you might loose another drive during the rebuild process. In all likely cases, since these cheap servers were brought & built in bulk, they all have drives from the same batch, so another drive failure shortly after the first drive failure is very likely.

Larger drives have a higher URE (Unrecoverable Read Error) then smaller drives, which means you'll have a failure sooner than later. The other two drives will take a lot more strain than it does normally and will produce errors much quicker than it should have.

Now you still need to worry about the replacement drive. Does the DC have one in stock, and can be be replaced quick enough, before one of the other drives fail as well? Let's just hope the drive set didn't break in Christmas Eve...

In all honesty, you're far better off with RAID1, or if you can convince them, RAID6 / RAID10.

RAID5 should only ever be used on a backup server which doesn't have much load and which won't cause unwanted downtime for your clients during a rebuilt period.

 

[Sinbad]

I would love to see you take out 1 drive (simulate a total drive failure) in a 3-drive RAID5 set, and still rebuild the RAID set. But, do this on a busy server with 3x 1TB HDD's. Forget about the nice theory about how RAID5 works for a second. Those RAID levels were designed with 36GB and 100GB HDD's in mind, many years ago. The rebuild will take about 18 hours, to two days, or more depending on the server's load. Chances are you're customer's will experience downtime as the server will take a performance hit. Even with a hardware RAID card, with BBU cache.
The next problem you sit with, is that you might loose another drive during the rebuild process. In all likely cases, since these cheap servers were brought & built in bulk, they all have drives from the same batch, so another drive failure shortly after the first drive failure is very likely.

Larger drives have a higher URE (Unrecoverable Read Error) then smaller drives, which means you'll have a failure sooner than later. The other two drives will take a lot more strain than it does normally and will produce errors much quicker than it should have.

Now you still need to worry about the replacement drive. Does the DC have one in stock, and can be be replaced quick enough, before one of the other drives fail as well? Let's just hope the drive set didn't break in Christmas Eve...

In all honesty, you're far better off with RAID1, or if you can convince them, RAID6 / RAID10.

RAID5 should only ever be used on a backup server which doesn't have much load and which won't cause unwanted downtime for your clients during a rebuilt period.

SO in fact, it's not RAID5 that is the problem, but in fact the risk of multiple drive failure.
Fact is, same scenario with RAID1 you'll be in the same boat... Rebuild time may not be 18 hours, but it will probably still be at least 6-12 hours - more if the server is busy.
RAID 6 just adds ANOTHER parity block to the stripe - but if the drives are as unreliable as you say, then if two drives fail you may as well expect 3 to fail... and I don't think RAID 6 will rebuild any faster than RAID 5 - not to mention the extra overhead in writes because of the extra parity block.

 

[SilverNodashi]

What you say applies, to people using PC based SATA RAID arrays.

I work for a large multinational. We STILL have more than 50% of our servers on ADAPTEC raid 5 with SCSI. Having worked for RAID 5 SCSI products fro more than 15 years. I have to say that there is a failure rate.

But on enterprise products. This is easily mitigated. Most of our server have RAID cards with memory backup.

The beauty these days is the Software interface notifies the syadm before there is a failure and we hot swap the drive. ALL of our servers has at least one Hard drive sitting as HOT Spare drive.

The usage of SATA drives for data redundancy is flawed because of the error rates of these drive increase when you pass a certain percentage of mechanical usage in a time frame.

Just to clear, this occurs on both SCSI drives and SATA raid arrays. But with enterprise scsi raid products , you are able to mitigate the errors leading to total loss of data.

You're not going to get a dedicated hosting server with 3x 1TB SCSCI HDD's, but rather with SATA HDD's and most providers use the cheaper servers, which often have onboard RAID without BBU CACHE.

Chances are also that the ISP won't be notified of a HDD error, but rather wait for the client to notify them when the server gives problems, so a HDD won't be swapped out while it's still in good shape, but rather when it has failed already.

RAID5 shouldn't be used in production, unless you manage your own servers, have a hot spare, and have smaller drives.

 

[Fulcrum29]

Raid is as worthy as the controller. I have seen controllers killing drives in array more than once. Speaking about probabilities...

 

[SilverNodashi]

SO in fact, it's not RAID5 that is the problem, but in fact the risk of multiple drive failure.
Fact is, same scenario with RAID1 you'll be in the same boat... Rebuild time may not be 18 hours, but it will probably still be at least 6-12 hours - more if the server is busy.
RAID 6 just adds ANOTHER parity block to the stripe - but if the drives are as unreliable as you say, then if two drives fail you may as well expect 3 to fail... and I don't think RAID 6 will rebuild any faster than RAID 5 - not to mention the extra overhead in writes because of the extra parity block.

I don't say drives are so unreliable. The industry has proven it....

RAID1 rebuilds different from RAID and the chances of loosing the 2nd drive while rebuilding the new one are much slimmer than loosing a 2nd drive in a 3drive RAID5 set while rebuilding the new drive.But hey, if you want to cheap out, go for it!

 

[Murmaider]

SO in fact, it's not RAID5 that is the problem, but in fact the risk of multiple drive failure.
Fact is, same scenario with RAID1 you'll be in the same boat... Rebuild time may not be 18 hours, but it will probably still be at least 6-12 hours - more if the server is busy.
RAID 6 just adds ANOTHER parity block to the stripe - but if the drives are as unreliable as you say, then if two drives fail you may as well expect 3 to fail... and I don't think RAID 6 will rebuild any faster than RAID 5 - not to mention the extra overhead in writes because of the extra parity block.

No, Raid 5 is inherently bad. In August 2012 Dell posted an advisory against the use of Raid 5 in any configuration - Their advisory states:

Finally, please be advised that the best practices for the use of RAID 5 and RAID 50 on Dell EqualLogic arrays have changed. The changes to the RAID policy best practice recommendations are being made to offer enhanced protection for your data.

RAID 5 is no longer recommended for any business critical information on any drive type
RAID 50 is no longer recommended for business critical information on Class 2 7200 RPM drives of 1TB and higher capacity

The issue is that if a URE is encountered during a Raid 5 rebuild (where its using parity and doing calculations to rebuild the missing data), it will breaks Raid 5 and the array. The URE would cause a 2nd drive in the raid 5 array to drop from the array, making it a 2 drive failure in a raid 5, which is unrecoverable.

If a URE is encountered during a Raid 1 rebuild, you should only lose that 1 sector to 1 raid stripe (worst case). This is because Raid 1/10 does not use Parity. The raid will continue to function.

 

[RNJudas]

Incorrect.

Anyway raid is not a backup. It's redundancy. A backup will allow you to recover a file that was deleted. Raid will not.

I tried to avoid it:

To see when it will come to the forums

 

[RNJudas]

Just to clear up confusion here - it looks like we had actually mis-labelled the product page - so thanks to everyone for bringing this issue up.

RAID has nothing to do with backups - as we don't actually offer backups as a general service. We do offer Free backup space (via FTP) on a backup server, but that is not something we perform - it's up to the client. The RAID label actually refers to RAID level (1 or 5) and not to anything related to backups.

We can offer R1Soft as a solution on request, which you can discuss with your account manager if you're looking for a solution like this.

Hope that helps clear up the confusion.

What.... No love for the heads up

BTW... Primary Backup Solution to RAID LEVEL is a typo of note.... .. Must have been a tough day at Afrihost

Though I am loving the discussion on raid with Sinbad and Dave... Heres my 5c:

RAID vs ZFS...

 

[Murmaider]

What.... No love for the heads up
Though I am loving the discussion on raid with Sinbad and Dave... Heres my 5c:

RAID vs ZFS...

Ah good ol ZFS, The way raid is meant to be.

 

[Xaze]

I'm posting the following open complaint due to lack of meaningful response from Afrihost support email,call centre, and no call back from a relevant department.

Dear Afrihost

I am the director of an online company which delivers online education and advice to its subscriber base. We have hosted our site for the past 18months with GoDaddy without any interruption of service.

This week we chose to migrate our site to Afrihost as we are focussed on the South African market and hosting in South Africa made sense.

However within 2 days you have suspended our domain and switched us off without any notification. Denying us access to our domains and database. When we call to ask what is happening we cannot get an answer other than it is to do with abuse.

This is not the quality of service we were expecting from Afrihost.
Due to your actions you have completely closed our business without any explanation or justification and deny us access to our clients and IPR.

I would like to know what you will do to rectify this extremely poor service in terms of compensation.

In speaking with Angel and others on your help desk, she said she did not see any cause for our site to be suspended for abuse so why has it been suspended?

I would appreciate hearing from Gian to understand how such an action could take place without any checks and balances on your part. Not only to switch off our service but to then have no point of escalation outside of normal business hours. This is very very poor on your part and I want to know what you will do to ensure it never happens again.

In short, I am livid.

Same happened to me. We had been housing afrihost for two years with no problems. But on day out of the blue our online store what closed. After investigation it turned out that someone at AH decided to list us as stammers. We use mailchimp for our newsletters and somehow an email was added to our list after an email campaign. We subsequent when through the listed and removed everyone that we was unsure about. But regardless our website was suspended without warning.

So we moved to Hetzner but sins moving most using afrihost can't access our site even though we have done everything in our power to play right.

As we are a online busines which relay on online campaigns it is harsh.

Mailchimp uses a double op in services so we are sure we are not spamming.

We have complained about this with AH with not result.

Kind regards Hannes

 

[AfriGuy]

Same happened to me. We had been housing afrihost for two years with no problems. But on day out of the blue our online store what closed. After investigation it turned out that someone at AH decided to list us as stammers. We use mailchimp for our newsletters and somehow an email was added to our list after an email campaign. We subsequent when through the listed and removed everyone that we was unsure about. But regardless our website was suspended without warning.

So we moved to Hetzner but sins moving most using afrihost can't access our site even though we have done everything in our power to play right.

As we are a online busines which relay on online campaigns it is harsh.

Mailchimp uses a double op in services so we are sure we are not spamming.

We have complained about this with AH with not result.

Kind regards Hannes

Sorry to hear about the poor experience here.
Just a heads up, it may be better to open a new thread considering this one's 2 years old

Either way, I'd gladly look at the query for you. If the mails were sent via Mailchimp then there isn't really a grounds for us to suspend your services unless the site breached our ToCs.
Shoot me a PM with the details and I'll gladly take a look at this for you.